Documente Academic
Documente Profesional
Documente Cultură
MATHEMATICS
MAL-523
METHODS OF APPLIED MATHEMATICS
1
MAL-523
METHODS OF APPLIED MATHEMATICS
Contents
Lesson No and Name
1 Fourier Transforms
2 Applications of Fourier Transforms
3 Curvilinear Co-ordinates
4 Random variable and Mathematical Expectation
5 Moments and Moment generating functions
6 Theoretical Discrete Distributions
7 Theoretical Continuous Distributions
8 Multiple and partial Correlation
Written by:
Prof .Kuldip Bansal
Department of Mathematics
G.J.University of Science & Technology
Hisar.
2
LESSON-1 FOURIER TRANSFORMS
Integral Transform: The integral transform of a function f(t) is defined by the equation
b
g(s) = ∫ f ( t ) k(s, t) dt, where k(s, t) is a known function of s and t, called the kernal of the
a
transform; s is called the parameter of the transform and f(t) is called inverse transform of g(s).
∫ | f (t ) | dt < ∞
−∞
3
∞
1
∫ f (t) e
−ist
f (s) = F[f ( t )] = F(s) = dt
2π −∞
Remarks:
∞
1
(i) If F(s) = ∫
2 π −∞
f ( t )e −ist dt, then
∫ F(s) e
ist
F −1 [F(s)] = ds
−∞
∞
1
∫ f (t) e
−ist
(ii) If F(s) = dt, then
2π −∞
∞
1
∫ F(s) e
ist
F −1 [F(s)] = ds
2π −∞
∞
1
∫ f (t) e
ist
(iii) If F(s) = dt, then
2π −∞
∞
1
∫ F(s) e
−ist
f(t) = ds
2π −∞
∞
2
π ∫ f (t ) sin st dt is called Fourier sine transform (FST) of f(t).
0
∞
2
The functions fc(t) =
π ∫ Fc (s) cos st ds
0
∞
2
fs(t) =
π ∫ Fs (s) sin st ds
0
are called the Inverse Fourier cosine & inverse Fourier sine transform of Fc(s) and Fs(s),
respectively.
4
Properties of Fourier Transform
(1) Linearity Property
If f1(t) and f2(t) are functions with Fourier Transform F1(s) and F2(s), respectively and C1
& C2 are constant, then Fourier Transform of C1 f1(t) + C2 f2(t) is
F[c1 f1(t) + c2 f2(t)] = c1 F[f1(t)] + c2 F[f2(t)]
= c1 F1(s) + c2 F2(s)
Proof: By definition
∞
1
∫e
−ist
L.H.S. = [c1 f1(t) + c2 f2(t)] dt
2π −∞
∞ ∞
c1 c2
∫ e −ist f1 ( t )dt + ∫e
−ist
= f 2 ( t )dt
2π −∞ 2π −∞
= C1 F[f1(t)] + C2 F[f2(t)]
= C1 F1(s) + C2 F2(s)
(2) Change of scale property or similarity theorem
If ‘a’ is a real constant and F(s) = F[f(t)] then
1 ⎛s⎞
F[f(at)] = F⎜ ⎟
|a| ⎝a⎠
Proof : For a > 0,
∞
∫ f (at)e
−ist
F[f(at)] = dt
−∞
1
Put at = x ⇒ dt = dx
a
∞ ⎛s⎞
1 −i ⎜ ⎟ x 1
∴ F[f(at)] =
2π
∫ f ( x )e
−∞
⎝a⎠
a
dx
∞ ⎛s⎞
1 1 −i ⎜ ⎟ t
=
a 2π ∫ f ( t )e ⎝a⎠
dt
−∞
1 ⎛s⎞
⇒ F[f(at)] = F⎜ ⎟
a ⎝a⎠
∞
1
∫ f (at)e
−ist
For a < 0 :- F[f(at)] = dt
2π −∞
5
dx
Put at = x ⇒ dt =
a
−∞ s
1 1 −i x
∴ F[f(at)] = .
a 2π ∫ f ( x )e a dx
∞
∞ s
−1 1 −i t
=
a
.
2π
∫ f (t) e a dt
−∞
−1 ⎛ s ⎞
⇒ F[f(at)] = F⎜ ⎟
a ⎝a⎠
1 ⎛s⎞
Hence, F[f(at)] = F⎜ ⎟
|a| ⎝a⎠
Particular case:- If a = −1, then
F[f(−t)] = F(−s)
Put t − u = v ⇒ dt = dv
∞
1
∫ f ( v) e
−is(v + u)
Θ F[f(t − u)] = dv
2π −∞
∞
1
∫ f ( v) e
−isu −isv
=e dv
2π −∞
∞
1
∫ f (t) e
−isu −ist
=e dt
2π −∞
= e−isu F[f(t)].
(4). Second shifting property
If F(s) = F[f(t)], then
F[f(t)eiat] = F(s − a)
6
∞
1
∫ f (t) e e−ist dt
iat iat
Proof :- F[f(t) e ] =
2π −∞
∞
1
∫ f (t) e
−i(s−a)t
= dt = F(s −a)
2π −∞
∫ F(s) e
ist
2π f ( t ) = ds
−∞
changing t to − t, we have
∞
∫ F(s)e
−ist
2πf (− t ) = ds
−∞
∫ F(t )e
−ist
2 π f (−s) = dt
−∞
∞
1
∫ F(t ) e
−ist
⇒ f(−s) = dt
2π −∞
⇒ f(−s) = F[F(t)]
⎧e − αt , t ≥ 0, α > 0
Example. 1. Find Fourier transform of f(t) = ⎨ …(1)
⎩ 0 , t<0
∞
1
∫ f (t ) e
−ist
Solution :- By definition, F(s) = dt
2π −∞
0 ∞
1 1
∫ f ( t )e −ist dt + ∫e
−αt −ist
⇒ F(s) = e dt
2π −∞ 2π 0
∞
1
∫e
−( α +ist ) t
= dt [using (1)]
2π 0
∞
1 ⎡ e −(α +is ) ⎤
⇒ F(s) = ⎢ ⎥
2π ⎣ − (α + is) ⎦ 0
7
1 ⎡ 1 ⎤ α − is
=
2π ⎢⎣0 + α + is ⎥⎦ = 2π (α 2 + s 2 )
⎧1 , | t |≤ a
Example.2. f(t) = ⎨
⎩0 , | t |> a
1 ⎡ ⎤
−a a ∞
⎢ ∫ f ( t )e dt + ∫ f ( t )e dt + ∫ f ( t )e dt ⎥
−ist −ist −ist
Solution F(s) =
2π ⎣⎢−∞ −a a ⎦⎥
= I1 + I 2 + I 3
−a
1
∫ f ( t )e
−ist
Let I1 = dt
2π −∞
Put t = −u ⇒ dt = −du
a
1
∴ I1 = ∫
2π ∞
f (−u ) eisu (−du)
∞
1
= ∫
2π a
f (− u )e isu du = 0
∞
Similarly I3 = ∫ f ( t )e −ist dt = 0
a
a
1
∴ I2 = ∫
2 π −a
f ( t )e −ist dt
a
1
= ∫
2 π −a
i. e −ist dt [Θ f(t) = 1]
a
1 ⎡ e −ist ⎤ −1
= ⎢ ⎥ = is [e −ias − e ias ]
2π ⎣ − i s ⎦ −a 2π
1 ⎛ e ias − e ias ⎞
⇒ I2 = ⎜ ⎟
2π ⎜⎝ is ⎟
⎠
1 2 sin as 2 ⎛ sin as ⎞
= . = ⎜ ⎟, s ≠ 0
2π is π ⎝ s ⎠
when s = 0,
8
2 a cos as
F(s) = as s→0 [By L-Hospital rule]
π (1)
2
= a.
π
⎧t, | t |≤ a ⎫
Example. 3 :- f(t) = ⎨ ⎬ …(1)
⎩0, | t |> a ⎭
∞
1
∫ f ( t )e
−ist
Solution :- F(s) = dt
2π −∞
1 ⎡− a a ∞ ⎤
⎢ ∫ f ( t )e dt + ∫ f ( t )e dt + ∫ f ( t )e dt ⎥
−ist −ist −ist
=
2π ⎢⎣ −a −a a ⎥⎦
a
1
∫t e
−ist
= dt [using (1)]
2π −a
1 ⎡⎛ e −ist ⎞
a a
e −ist ⎤
= ⎢⎜ t ⎟ − ∫ (1) dt ⎥
2π ⎢⎜⎝ − i s ⎟
⎠ −a −a − is ⎥
⎣ ⎦
1 ⎡ 2a cos as 1 ⎤
⇒ F(s) = ⎢ − 2 si sin as⎥
2 π ⎣ − is s ⎦
2
= 2
[i as cos as − i sin as]
s 2π
2 i
⇒ F(s) = [as cos as − sin as]
π s2
Example. 4. If F(s) = F[f(t)]. Find F.T. of f(t) cos at
Solution :- We know that
1 iat
cos at = (e + e −iat )
2
1 1
Then F[f(t) cos at] = F[f(t) eiat] + F[f(t) e−iat]
2 2
9
or prove by definition [By linearity property]
1 1
⇒ F[f(t) cos at] = F(s −a) + F(s + a) [By using shifting property]
2 2
Example. 5. If Fs(s) and Fc(s) are FST and FCT of f(t) respectively, then
1
Fs[f(t) cos at] = [Fs(s + a) + Fs(s − a)]
2
1
Fc[f(t) cos at] = [Fc(s + a) + Fc(s − a)]
2
1
Fs[f(t) sin at] = [Fc(s − a) − Fc(s + a)]
2
1
Fc[f(t) sin at] = [Fs (s + a) − Fs(s −a)]
2
Solution :- (i) By definition of FST,
∞
2
Fs[f(t) cos at] =
π ∫ f (t ) cos at sin st dt
0
∞
1 2
= .
2 π ∫ f (t ) 2 cos at sin st dt
0
1⎡ 2 ⎤
∞ ∞
2
= ⎢
2 ⎣⎢ π ∫ f ( t ) sin( s + a ) t dt +
π ∫0
f ( t ) sin( s − a ) t dt ⎥
⎥⎦
0
1
= [Fs (s + a) + Fs (s −a)]
2
(ii) By definition FCT,
∞
2
Fc[f(t) cos at] =
π ∫ f (t ) cos at cos st dt
0
∞
1 2
=
2 π ∫ f (t ) [cos(s + a)t + cos (s−a) t]dt
0
10
1⎡ 2 ⎤
∞ ∞
2
= ⎢
2 ⎣⎢ π ∫ f (t ) cos(s + a )t dt + π ∫ f ( t ) cos(s − a ) t dt ⎥
0 0 ⎦⎥
1
= [Fc (s + a) + Fc(s −a)]
2
(iii) By definition of FST,
∞
2
Fs[f(t) sin at] =
π ∫ f (t ) sin at sin st dt
0
∞
1 2
=
2 π ∫ f (t ) [cos(s −a) − cos (s + a)] dt
0
1⎡ 2 ⎤
∞ ∞
2
= ⎢
2 ⎢⎣ π ∫ f (t ) cos(s − a )dt − π ∫ f (t ) cos(s + a )dt ⎥⎥
0 0 ⎦
1
= [Fc (s−a) − Fc(s + a)]
2
(iv) By definition of FCT,
∞
2
Fc[f(t) sin at] =
π ∫ f (t ) sin at cos st dt
0
∞
1 2
= .
2 π ∫ f (t ) [sin (s + a)t − sin (s−a)t] dt
0
1⎡ 2 ⎤
∞ ∞
2
= ⎢
2 ⎢⎣ π ∫ f (t ) sin(s + a )t − π ∫ f ( t ) sin( s − a ) t dt ⎥
⎥⎦
0 0
1
= [Fs (s + a) − Fs(s + a) − Fs(s − a)].
2
∫e
−at
Let cos st dt = I1 …(1)
0
11
∞
∫e
−at
sin st dt = I2 …(2)
0
1 s
= − I2 …(3)
a a
Integrating (2) by parts, we have
s
I2 = I1 …(4)
a
⎡ ⎡ −1 ⎤
∞
s
∞ ⎤
⎢Θ I 2 = ⎢ e −at sin st ⎥ + ∫ e −at cos st dt ⎥
⎢⎣ ⎣a ⎦0 a 0 ⎥⎦
solving (3) and (4) for I1 & I2,
1 s s
I1 = − . I1
a a a
⎛ s2 ⎞ 1 1 a2
⇒ I1 ⎜⎜1 + 2 ⎟⎟ = ⇒ I1 = . 2
⎝ a ⎠ a a s + a2
a
⇒ I1 =
s + a2
2
s a s
and I2 = . 2 2
= 2
a s +a s + a2
2 2 a
Hence Fc [e−at] = I1 = …(5)
π π s + a2
2
2 2 s
and Fs[e−at] = I2 = …(6)
π π s + a2
2
2 ⎡ (s 2 + a 2 )1 − a.2a ⎤
Fc[−t e−at] = ⎢ ⎥
π ⎣ (s 2 + a 2 ) 2 ⎦
12
2 ⎡ − a 2 + s2 ⎤
⇒ − Fc[t e−at] = ⎢ 2 2 2⎥
π ⎣ (s + a ) ⎦
2 ⎡ a 2 − s2 ⎤
⇒ Fc[ t e−at] = ⎢ 2 2 2⎥
π ⎣ (s + a ) ⎦
Differentiating both sides of (6) w.r.t. ‘a’, we get
2 ⎡ (s 2 + a 2 )0 − s. 2a ⎤
Fs [− t e−at] = ⎢ 2 2 2 ⎥
π ⎣ (s + a ) ⎦
2 2as
⇒ −Fs[ t e−at] = −
π (s + a 2 ) 2
2
2 2as
⇒ Fs [ t e−at] = .
π (s + a 2 ) 2
2
2 s2 2 s
Fs[e−t] = =
π s2 +1 π s2 +1
2 1
Fc[e−t] = . 2
π s +1
Results:
e αt
∫ e cos βt dt =
αt
(α cos βt + β sin βt)
α2 + β2
e αt
∫e
αt
sin βt dt = 2 (α sin β t − β cos βt)
α + β2
2
Example. 7 :- Find F. T. of f(t) = e −t /2
⎡ − t2 ⎤ ∞ t2
1 −
Solution :- By definition, F⎢e 2 ⎥ = ∫e 2 e−ist dt
⎢ ⎥ 2π
⎣ ⎦ −∞
∞ 1 −s 2
1 − ( t +is ) 2
−t 2 / 2
⇒ F[e ]=
2π
∫ e 2 .e 2 dt
−∞
2 ∞ 1
e −s / 2 − ( t +is ) 2
=
2π
∫ e 2 dt
−∞
13
1
Put (t + is) = y ⇒ dt = 2 dy
2
2 ∞
−t 2 / 2 e −s / 2 2
∫ e . 2 dy
−y
⇒ F[e ]=
2π −∞
2 2
/2 ∞
e −s − y2 e −s /2
=
π
∫ e dy =
π
. π
−∞
2 2 2 2
⇒ F [e − t /2
] = e −s /2
, Fc [e− t /2
] = e −s /2
2
Example. 8. Find Fourier cosine transform of f(t) = e − t
∞
2 2 −t 2
Solution :- Fc [e − t ] =
π ∫e cos st dt = I …(1)
0
∞
1 2 −t 2
= .
2 π ∫ ( −2 t e ) sin st dt
0
Integrating by parts,
dI 1 2 ⎡ ⎤
∞ ∞
−t 2 2
= . ⎢sin ste − s ∫ cos st. e − t dt ⎥
ds 2 π ⎢⎣ 0 0 ⎥⎦
∞
s 2 −t 2
=−
2 π ∫e cos st dt
0
dI s
⇒ =− I [using (1)]
ds 2
dI − s
⇒ = ds
I 2
s2
Integrating, log I = − + log A
4
2
⇒ I = A e−s /4
…(2)
14
∞
2 2
when s = 0, from (1), I =
π ∫ e − t dt
0
2 π 1
⇒ I= . = …(3)
π 2 2
Also when s = 0, from (2) ⇒ I = A …(4)
1
∴ from (3) & (4), we get A =
2
1 −s2 / 4
∴ (2) gives I = e
2
2
1⎛ s ⎞
2 2 1 1 − 4 ⎜⎝ a ⎟⎠
Extension :- Fc [ e −a t
]= . e
2 a
−s 2
−a 2 t 2 1 2 1 ⎛s⎞
⇒ Fc [e ]= e 4a [using change of scale probability F[f(at)]= F⎜ ⎟
a 2 |a| ⎝a⎠
1
If a = , we have
2
2 2
Fc [e − t /2
] = e −s /2
Self-reciprocal function:
A function f(t) with the property that F[f(t)] = f(s) is said to be self-reciprocal under
2
Fourier transform, e.g. the function e − t / 2 is self-reciprocal function under F.T. The function
2
e − t / 2 is also self reciprocal under F.C.T.
2 2
To prove :- Fc [e − t /2
] = e −s /2
2 2
Let Fc [e − t /2
] = e −s /2
∞
2 −t 2 / 2 2
∫e cos st dt = e −s /2
⇒
π 0
2
Hence the function t e −t /2
is self-reciprocal function under Fourier Sine Transform.
15
⎧1 , | t |≤ a
Example. :- We know that if f(t) = ⎨
⎩0 , | t |> a
2 sin sa
Then F(s) = ,s≠0 …(1)
π s
∞
1
By definition of F. T., F(s) =
2π
∫ f(t) e−ist dt
−∞
∞
1
∫ F(s)e
−st
then f(t) = ds
2π −∞
∞
1 sin sa ist ⎧1 , | t |< a
⇒ f(t) = ∫
π −∞ s
e ds = ⎨
⎩0 , | t |> a
∞
1 sin sa
π −∫∞ s
L.H.S. = [cos st + i sin st] ds
∞ ∞
1 sin sa i sin sa
= ∫
π −∞ s
cos st ds + ∫
π −∞ s
sin st ds
∞ ⎧π , | t |< a
sin sa ⎪
⇒ ∫ s cos st ds = ⎨0 , | t |> a
−∞ ⎪π / 2 , | t |= a
⎩
∞ ⎧π / 2 , | t |< a
sin sa cos st ⎪
or ∫ s
ds = ⎨0 , | t |> a …(2)
0 ⎪π / 4 , | t |= a
⎩
∞
sin s
Evaluate ∫ s
ds
0
16
∞
sin s π
∫ s
ds = when |t| < a.
2
0
∞
1
∫e e−xt φ(t) dt
−ist
=
2π 0
∞
1
∫e
−( x +is ) t
= φ(t) dt
2π 0
∞
1
= ∫
2π 0
e −pt φ(t) where p = x + i
1
⇒ F[f(t)] = L[φ(t)]
2π
F.T. of Derivatives
If F[f(t)] = F(s) & f(t)→0 as t→± ∞, then
F[f ′(t)] = i s F[f(t)] = i s F(s)
Proof :- By definition,
∞
1
∫ f ' ( t )e
−ist
F[f′(t)] = dt
2π −∞
∞
=
1
2π
[ ∞
e −ist ( t ) −∞ −
1
2π
](−is) ∫ f ( t )e −ist dt
−∞
∞
1
∫ f ( t )e
−ist
= is . dt [Θ f(t)→0 as t→± ∞]
2π −∞
17
F[f ′′(t)] = is F[f′(t)] = (is)2 F[f(t)]
In general, we have
F[f(n)(t)] = is F[f(n−1)(t)]
= (i s)n F[f(t)] = (is)n F(s)
Find Fourier sine & cosine transform of f′(t), f′′(t).
Derivation :- By definition
∞
2
Fc[f′(t)] =
π ∫ f ' (t ) cos st dt
0
∞
2 2
=
π
[cos st f ( t )]∞0 + s
π ∫ f (t ) sin st dt
0
2
⇒ Fc[f ′(t)] = f(0) + s Fs [f(t)] …(1)
π
∞
2
Fs[f ′(t)] =
π ∫ f ' (t ) sin st dt
0
∞
2
= [sin stf ( t )]∞0 − s 2 ∫ f (t ) cos st dt
π π 0
2
Fc[f ′′(t)] = − f′(0) + s Fs [f′(t)] (using (1)
π
2
= −f′(0) + s[−s Fc[f(t)]] [By using (2)]
π
2
⇒ Fc[f ′′(t)] = −f′(0) − s2 Fc[f(t)] …(3)
π
and Fs[f ′′(t)] = −s Fc[f′(t)] [using (2)]
18
⎧ 2 ⎫
= − s ⎨− f (0) + s Fs [f ( t )]⎬ [using (1)]
⎩ π ⎭
2
⇒ Fs[f ′′(t)] = s f(0) − s2 Fs[f(t)] …(4)
π
Theorem:- If F[f(t)] = F(s), then
dn
n
F[f(t)] = (−i)n F[tn f(t)], n = 1, 2, 3
ds
Proof:- By definition,
∞
1
∫ f (t) e
−ist
F[f(t)] = F(s) = dt
2π −∞
∞
1
∫ f (t ) (−i t) e
−ist
= dt
2π −∞
∞
d 1
∫ t f ( t )e
−ist
⇒ F(s) = (−i) dt
ds 2π −∞
d
⇒ F(s) = (−i) F[t f(t)]
ds
∞
d2 1
∫ (−i t ) f(t) e−ist dt
2
Now 2
F(s) =
ds 2π −∞
∞
1
= (−i)2 ∫t f(t) e−ist dt
2
2π −∞
d2
⇒ 2
F(s) = (−i)2 F[t2 f(t)]
ds
Generalising the result, we have
dn
F[f ( t )] = (−i) n F[tn f(t)], n = 1, 2,…
ds n
Theorem :- If F[f(t)] = F(s)
19
∞
and ∫ f (t )dt = F(0) = 0
−∞
⎡t ⎤ 1
then F ⎢ ∫ f ( x )dx ⎥ = F(s)
⎣⎢−∞ ⎦⎥ i s
t
Proof : Consider φ(t) = ∫ f (x)dx
−∞
F(s)
⇒ Φ(s) =
is
F(s)
⇒ F[φ(t)] =
is
⎡t ⎤ 1
⇒ F ⎢ ∫ f ( x )dx ⎥ = F(s)
⎢⎣−∞ ⎥⎦ i s
Convolution :- Let f1(t) and f2(t) be two given functions, the convolution of f1(t) & f2(t) is
defined by the function
∞
1
f(t) =
2π
∫ f1 (x) f2(t − x)dx
−∞
= f1(t) ∗ f2(t)
Special case :-
f1(t) = 0 for t < 0
f2(t) = 0 for t < 0
Then
t
1
f(t) = f1(t) ∗ f2(t) =
2π
∫ f1 (x ) f2(t − x) dx
0
1 ⎡ ⎤
0 t ∞
Proof :- f(t) = ⎢ ∫ f1 ( x )f 2 ( t − x )dx + ∫ f1 ( x )f 2 ( t − x )dx + ∫ f1 ( x )f 2 ( t − x )dx ⎥
2π ⎣⎢−∞ 0 t ⎦⎥
20
…(1)
Now I1 = 0 [Θf1(t) = 0 for t < 0] …(2)
∞
1
and I3 =
2π
∫ f1 (x )f 2 (t − x )dx
t
Commutative Property:
f1(t) ∗ f2(t) = f2(t) ∗ f1(t)
Proof :- By definition,
∞
1
f1 ∗ f2 =
2π
∫ f1 (x ) f2(t − x)dx
−∞
put t − x = y ⇒ dx = − dy
when x = −∞, y = ∞
and when x = ∞, y = −∞
−∞
1
⇒ f1 ∗ f2 =
2π
∫ f1(t − y) f2(y) (−dy)
∞
∞
1
⇒ f1 ∗ f2 =
2π
∫ f 2 ( y) f1(t − y) dy = f2 ∗ f1
−∞
Associativity property
(f1 ∗ f2) ∗ f3 = f1 ∗(f2 ∗ f3) …(1)
Take g(t) = f1 ∗ f2
h(t) = f2 ∗ f3
Then (1) becomes, g(t) ∗ f3 = f1 ∗ h(t)
21
Convolution Theorem (or Falting Theorem) on Fourier Transforms
If F[f1(t)] = F1(f)
F[f2(t)] = F2(s) i.e.
F[f1(t) ∗ f2(t)] = F1(s) F2(s)
where
∞
1
f1(t) ∗ f2(t) =
2π
∫ f1 ( x) f2(t − x)dx
−∞
1
∞ ⎡ 1 ∞ ⎤ −ist
=
2π
∫ ⎢⎢ 2π ∫1 2
f ( x ) f ( t − x ) dx ⎥ e dt
−∞ ⎣ −∞ ⎦⎥
[By using definition of convolution]
changing the order of integration, we get
1
∞ ⎡ 1 ∞ ⎤
∫ 1 ⎢⎢ 2π ∫2
−ist
F[f1(t) ∗ f2(t)] = f ( x ) f ( t − x ) e dt ⎥ dx
2π −∞ ⎣ −∞ ⎦⎥
1
∞ ⎡ 1 ∞ ⎤
∫ 1 ⎢⎢ 2π ∫2
−is ( t − x ) −isx
= f ( x ) f ( t − x ) e e dt ⎥ dx
2π −∞ ⎣ −∞ ⎦⎥
1
∞ ⎡ 1 ∞ ⎤ −isx
∫ ∫
−isy
= f 1 ( x ) ⎢ f 2 ( y ) e dy ⎥ e dx
2π −∞ ⎣⎢ 2π −∞ ⎦⎥
[By putting t− x = y ⇒ dt = dy]
∞
1
∫ f1 ( x ) e
−isx
⇒ F[f1(t) ∗ f2(t)] = F2 (s)dx
2π −∞
∞
1
∫ f1 ( x ) e
−isx
= F2(s) dx
2π −∞
22
By definition of F.T., we have
∞
1
∫ f1 ( x ) f 2 ( x ) e
−isx
F[f1(t) f2(t)] = dx
2π −∞
1
∞ ⎡ 1 ∞ ⎤
∫ f1 (x)⎢⎢ 2π ∫ F2 (s' )e ds'⎥ e−isx dx
is 'x
F[f1(t) f2(t)] =
2π −∞ ⎣ −∞ ⎦⎥
1
∞ ⎡ 1 ∞ ⎤
∫ ds' F2 (s' )⎢⎢ 2π ∫ f1 ( x ) e
−i (s −s ') x
= dx ⎥
2π −∞ ⎣ −∞ ⎦⎥
∞
1
⇒ F[f1(t) f2(t)] =
2π
∫ F2 (s' )F1 (s − s' )ds'
−∞
∞
1
∫ F(z) e
−izt
Then inverse function f(t) = dz
2π −∞
Example :- Find F.T. of f(t) = e−|t| and verify the inverse transform.
Sol. Now |t| = t for t > 0
|t| = −t for t < 0
So
⎧⎪e t , t<0
f(t) = ⎨
⎪⎩e − t , t>0
∞
1
∫ f ( t )e
−izt
∴ F(z) = dt
2π −∞
1 ⎡ ⎤
0 ∞
⎢∫ e dt + ∫ e − t (1+iz ) dt ⎥
t (1−iz )
=
2π ⎣⎢−∞ 0 ⎦⎥
23
1 ⎡ 1 1 ⎤ 1 2
= ⎢1 − i z + 1 + i z ⎥ =
2π ⎣ ⎦ 2π (1 + z 2 )
We can invert the transform using contour integration (Residue theorem). First consider t > 0,
then
∞ ∞
1 1 2 1 e izt
∫ e dz = ∫
izt
f(t) = dz
2π −∞ 2π 1 + z 2 π −∞ 1 + z 2
1 e izt
= lim
π R →∞ ∫ 1 + z 2 dz
C
C
z=i
−R O R
(z − i)e izt 1
There is a simple pole at z = i and Res (z = i) = Lim = e−t
z→i (z − i)(z + i) 2i
Therefore by Residue theorem,
∞
1 e izt 1 ⎛ e −t ⎞
∫
π −∞ 1 + z 2
dz =
π
2 πi ⎜
⎜
⎝ 2i ⎠
⎟ = e−t
⎟
Consider t < 0, then we choose the contour with a semi-circular arc lying below the x-axis where
c is the contour shown in figure.
−R O R
x
z=i
(z + i)e izt et
Res (z = −) = Lim =
z →−i (z + i)(z − i) − 2i
24
But in this case contour is in clockwise direction, hence the desired result is
∞
1 e izt −1 ⎛ − et ⎞
∫
π −∞ 1 + z 2
dz =
π
2 πi⎜⎜
⎝ 2i ⎠
⎟ = et
⎟
=1
⎧sin wt , 0 < t < ∞
Example : Find F.T. of f(t) = ⎨
⎩ 0 , otherwise
or f(t) = sin wt H(t)
where
⎧1, t>0
H(t) = ⎨
⎩0, t<0
is a unit step function or Heaviside’s unit step function.
1 ∞
Sol. Now F(z) = ∫
2π 0
sin wt e−izt dt
1 1⎡ 1 1 ⎤
= . ⎢ − ⎥
2 π 2i ⎣ i(z − w ) i( w + z) ⎦
−1 ⎡ −1 1 ⎤ 1 ⎡− z − w + z − w ⎤
= + =
2i 2 π ⎣ z − w z + w ⎦ 2 2 π ⎢⎣
2 ⎢ ⎥ z2 − w 2 ⎥⎦
−1 ⎛ w ⎞
= ⎜ ⎟
2π ⎝ z 2 − w 2 ⎠
25
−w
Here F(z) = is analytic in z-plane except at z = ± w
2π (z 2 − w 2 )
−1 we izt
If t > 0 :- f(t) = Lim ∫ 2 dz
2 π R →∞ C z − w 2
where c is contour.
There are two simple pole z = w & z = w inside c.
y
z = −w z=w
x
−R R
1 ∞
where f(t) = ∫
2 π −∞
F(z)e izt dz
1 ∞ −w
= ∫
2 π −∞ 2 π ( z − w )
2 2
eizt dz
− 1 ∞ we izt
2 π −∫∞ z 2 − w 2
⇒ f(t) = dz
−1 ∞ 1 we izt −1 we izt
∴ ∫
2 π −∞ 2 π z 2 − w 2
dz =
2π R →∞ C∫ z 2 − w 2
Lim dz
weizt weiwt
Res (z = w) = Lim (z − w ) =
z→w (z − w )(z + w ) 2w
e iwt
=
2
weizt we −iwt
Res (z = −w) = Lim (z + w ) =
z →− w (z − w )(z + w ) − 2w
− 1 −iwt
= e
2
Therefore by Residue theorem,
26
−1 we izt −1
Lim ∫ 2 dz = (2 πi) (Sum of residue
Cz −w
2
2π R → ∞ 2π
⎡ e iwt − e iwt ⎤
= − i⎢ ⎥
⎣ 2 ⎦
⇒ f(t) = −i(i sin wt) = sin wt
If t < 0, we choose the contour with a semi-circular arc below x-axis & since there are no poles
inside the contour, the result is zero.
27
LESSON-2 APPLICATIONS OF FOURIER TRANSFORMS
Solution of Ordinary Differential Equation
Consider the nth order differential equation
dn y d n −1 y dy
An n
+ a n −1 n −1
+ .... + a 1 + ay = f(t)
dt dt dt
Taking F.T. of both sides,
[an (i s)n + an−1 (i s)n−1 +…+ a1(i s) + a] F[y(t)] = F[f(t)]
Let F[f(t)] = G(s)
F[y(t)] = Y(s)
P(i s) = an (i s)n + an−1 (i s)n−1 +….+ a1(i s) + a
is a polynomial in (i s).
G (s )
Then we have Y(s) =
P (i s )
Taking inverse F.T., we have the solution is given by
1 ∞
y(t) = ∫
2 π −∞
Y(s) eist ds
d2y dy
2
+ 3 + 2 y = e −|t| …(1)
dt dt
Solution :- Taking F.T. on both sides,
[(i z)2 + 3(i z) + 2] F[y(t)] = F[e−|t|]
1 2
⇒ P(i z) F[y(t)] =
2π 1 + z 2
1 2
⇒ Y(z) = where Y(z) = F[y(t)]
2π (1 + z 2 ). 1
P(i z)
1 ∞ 2 e izt dz
⇒ y(t) = ∫
2 π −∞ 2 π (1 + z 2 )[(i z) 2 + 3iz + 2]
1∞ e izt dz
= ∫ 2
π −∞ (z + 1)(− z 2 + 3iz + 2)
28
−1 ∞ e izt dz
π −∫∞ (z 2 + 1)(z 2 − 3iz − 2)
=
−1 ∞ e izt dz
π −∫∞ (z − 1)(z + 1)(z − i)(z − 2i)
=
−1 ∞ e izt dz
π −∫∞ (z + i)(z − i) 2 (z − 2i)
=
For t > 0
The singularity within the contour are a simple pole at z = 2i and a double pole at z = i.
(z − 2i)e izt
∴ Res. (z = 2i) = Lim
z →2 i ( z + i)( z − i ) 2 ( z − 2i )
e izt z = 2i
= Lt C
z →2 i ( z + i)( z − i) 2
z=i
−2 t
e = 1 −2t
= 2
= e
3i .i 3i R
z=i
d (z − i) 2 e izt
Res. (z = i) = Lim
z→i dz ( z + i )( z − i ) 2 ( z − 2i )
d e izt
= Lim
z→i dz (z + i)(z − 2i)
2i.(−i)(i t )e − t − e − t i
=
4i 2 .(−i) 2
2i te − t − i e − t 2i t e − t − i e − t
= =
4(−1)(−1) 4
−1 ⎡ + i −2 t 2i t e − t − i e − t ⎤
∴ y(t) = .2 πi ⎢ e + ⎥
π ⎣ 3(1) 4 ⎦
29
⎡ 4ie −2 t + 6i t e − t − 3 i e − t ⎤
= − 2i ⎢ ⎥
⎣ 12 ⎦
−i
= [6 i t e−t − 3i e−t + 4 i e−2t]
6
1
= (6 t e−t − 3 e−t + 4 e−2t)
6
2 −2t 1 −t
= e + t e −t − e …(2)
3 2
Verification :- Put (2) in L.H.S. of DE. (1) we get
d 2 y 3dy d ⎡ 2 −2 t −t −t 1 −t ⎤
L.H.S. = + + 2 y = e ( −2) + e − t e + e
dt 2 dt dt ⎢⎣ 3 2 ⎥⎦
⎡ − 4 −2 t 1 ⎤ 4
+ 3⎢ e + e − t − t e − t + e − t ⎥ + e −2 t + 2 t e−t − e−t
⎣ 3 2 ⎦ 3
− 4 −2 t − 3 −t 9
L.H.S. = e ( −2 ) e − e − t + t e − t − 4e − 2 t + e − t − 3 t e − t
3 2 2
4 −2 t
+ e + 2 t e − t − e − t = 2 e − t − e − t = e −t
3
For t < 0
(z + i)e izt
Res. (z = −i) = Lim
z →− i ( z + i)( z − i ) 2 ( z − 2i )
et et et
= = =
(−2i) 2 (−3i) 4(−1)(−3i) 12i
−1 et et
So y(t) = (−2 πi) = …(3)
π 12i 6
Verification :- Put (3) in L.H.S. of (1),
d 2 y 3dy d ⎡et ⎤ e t 2e t
L.H.S. = + + 2 y = ⎢ ⎥ + 3 +
dt 2 dt dt ⎣ 6 ⎦ 6 6
e t 3e t 2e t 6e t
= + + = = e t . Hence verified.
6 6 6 6
30
Example:- Solve using F.T. techniques
d 2 y 3dy
+ + 2 y = H(t) sin wt (1)
dt 2 dt
Solution :- Then we have on taking F.T. on both sides,
1 w
[(i z)2 + 3 i z + 2] F[y(t)] =
2π w − z 2
2
1 ∞ 1 ⎛ w ⎞ e izt dz
Then y(t) = ∫ ⎜ ⎟
2 π −∞ 2 π ⎝ w 2 − z 2 ⎠ (− z 2 + 3iz + 2)
w ∞ e izt dt
2 π −∫∞ (z 2 − w 2 )(z 2 − 3iz − 2)
⇒ y(t) =
w ∞ e ιzt dt
2π −∫∞ (z 2 − w 2 )(z − i)(z − 2i)
=
For t > 0
In this case singularity are at z = ± w, z = 2i, z = i and all these lies inside the upper half
plane.
⎡ (z − 2i) e izt ⎤
Res. (z = 2i) = Lim ⎢ ⎥
z →2i ( z − w )( z + w )( z − i )(z − 2i)
⎣ ⎦
e −2 t e −2 t i e −2 t
= = =
(2i − w )(2i + w )i − (4 + w 2 )i 2 w 2 + 4
⎡ (z − i) e izt ⎤
Res. (z = i) = Lim ⎢ ⎥
z→i ( zi −)( z − 2i )( z 2 − w 2 )
⎣ ⎦
e−t e−ti e−t
= = =
− i(−1 − w 2 ) (−1 − w 2 ) i( w 2 + 1)
(z − w ) e izt
Res. (z = w) = Lim
z→w (z − w )(z + w )(z − i)(z − 2i)
e iwt e iwt
= =
2 w ( w − i)( w − 2i) 2 w ( w 2 − 3iw − 2)
(z + w ) e izt
Res. (z = −w) = Lim
z →− w ( z − w )(z + w )(z − i )(z − 2i )
31
e −iwt e −iwt
= =
(−2w )(− w − i)(− w − 2i) − 2w ( w + i)( w + 2i)
e −iwt − e −iwt
⇒ Res. (z = −w) = =
− 2 w ( w 2 + 3iw − 2) 2 w ( w 2 + 3iw − 2)
− 4 we −2 t w e−t iw 2 e −iwt
y′′ + 3y′ + 2y = + 2 +
w2 + 4 w + 1 2( w + i)( w + 2i)
e −iwt
⇒ y′′ + 3y′ + 2y = [ i w2 − 3w − 2i]
2( w + i)( w + 2i)
e iwt
+ [−i w2 − 3w + 2i]
2( w − i)( w − 2i)
32
⇒ y′′ + 3y′ + 2y = H(t) sin wt
Hence y′′ + 3y′ + 2y = sin wt. H(t) is verified for t > 0.
for t < 0, there is no pole inside the lower half plane so residue = 0.
∴ y(t) = 0
∴ y′′ + 3y′ + 2y = H(t) sin wt is verified.
For t < 0.
∴ for t < 0 , H(t) = 0
So R.H.S. of (1) is zero.
And equal to L.H.S. Hence verify the result.
e − at
Example: Find FST of
t
e − at
Solution :- The sine transform of function f(t) =
t
2∞
π ∫0
I = Fs[f(t)] = f ( t ) sin st dt
2 ∞ e − at
π ∫0 t
= sin st dt …(1)
∞
2 ⎡ e −at ⎤
= ⎢ 2 ( − a cos st + s sin st ) ⎥
π ⎣ a + s2 ⎦0
d 2 a dI
⇒ Fs [f ( t )] = . 2 =
ds π a +s 2
ds
Integrating we get,
−d 2 a
∫ ds Fs [f ( t )] = ∫
π a + s2
2
ds
2 ⎛s⎞
⇒ I = Fs[f(t)] = tan −1 ⎜ ⎟ + A …(2)
π ⎝a⎠
33
where A is constant of integration. For s = 0, we get from (2),
2
I= (0) + A ⇒ I = A …(3)
π
2∞
π ∫0
when s = 0, from (1), I = (0) dt = 0 …(4)
⎡ e − at ⎤ 2 ⎛s⎞
Fs ⎢ ⎥= tan −1 ⎜ ⎟ .
⎣ t ⎦ π ⎝a⎠
Solution of partial differential equations (boundary value problems)
Example: - Determine the distribution of temperature in the semi-infinite medium, x ≥ 0 when
the end x = 0 is maintained at zero temperature & the initial distribution of temperature is f(x).
Solution :-
−∞ 0 ∞
∂u ( x , t ) 2 ∂ u ( x, t )
2
=c , x > 0, t > 0 …(1)
∂t ∂x 2
where u(x, t) is the distribution of temperature at any point x and time t. We want to determine
the solution of (1) subject to initial condition u(x, 0) = f(x) …(2)
and the boundary condition u(0, t) = 0 …(3)
Since (u)x=0 is given, we apply F.S.T.
Denote Fs[u(x, t)] = u s (s, t ) = u s and Fc[u(x, t)] = u c
Taking FST of both sides of (1), we get
2 ∞ ∂u 2 ∞ ∂ 2u
∫
π 0 ∂t
sin st dx = c2
π ∫0 ∂x 2
sin sx dx
34
2 d∞ 2 ∞ ∂ 2u
⇒ ∫
π dt 0
u sin s x dx = c2
π ∫0 ∂x 2
sin sx dx
2 d∞ d
⇒ ∫
π dt 0
u sin sx dx =
dt
Fs[u(x, t)]
⎧⎪ 2 ⎡ ∂u ⎤
∞
2 ∞ ∂u ⎫⎪
π ∫0 ∂x
=c ⎨ 2
sin sx − s cos sx dx ⎬
⎪⎩ π ⎢⎣ ∂x ⎥⎦
0 ⎪⎭
d 2 ∞ ∂u ∂u
us = 0 − c s ∫
2
⇒ cos sx dx if → 0 as x→∞
dt π 0 ∂x ∂x
2 ∞ ∂u
π ∫0 ∂x
2
=−C s cos sx dx
∞
2
= − c2s [u ( x, t ) cos sx ]0∞ − c 2 s 2 2 ∫ u sin sx dx
π π 0
d 2
us = c2 s u (0, t ) − c 2 s 2 u s ; assuming u→0 as x→∞
dt π
du s ⎡ 2 ⎤
⇒ = c 2 ⎢ s u (0, t ) − s 2 u s ⎥
dt ⎣ π ⎦
By using (3), u(0, t) = 0, we get
d
⇒ u s + c 2s 2 u s = 0 …(4)
dt
Also taking FST of (2), we get
Fs[u(x, 0)] = Fs[f(x)]
⇒ u s (s, 0) = f s (s) …(5)
35
⇒ A = f s (s) [using (5)]
Hence solution is
2 2
u s (s, t ) = f s (s)e −c s t
2∞
π ∫0
⇒ u(x, t) = u s sin sx ds
2 ∞ 2 2
⇒ u(x, t) = ∫
π 0
f s (s)e −c s t sin sx ds
∂u ∂ 2u
=K 2 …(1)
∂t ∂x
subject to conditions
(i) u = 0 when t = 0 , x ≥ 0
i.e. u(x, 0) = 0
∂u
(ii) = −μ (a constant) when x = 0, t > 0
∂x
∂u
i.e. (0, t) = −μ
∂x
or ux(0, t) = −μ
⎛ ∂u ⎞
Solution :- Since ⎜ ⎟ is given,
⎝ ∂x ⎠ x =0
So taking Fourier cosine transform of both sides of (1),
2 ∞ ∂u 2 ∞ ∂ 2u
π ∫0 ∂t π ∫0 ∂x 2
cos sx dx = K cos sx dx
∞
d 2 ⎡ ∂u ⎤ 2 ∞ ∂u
⇒
dt
u c (s, t ) = K
π ⎢⎣ ∂x
cos sx ⎥⎦ + K s
π ∫ ∂x sin sx dx
0 0
36
d 2 ⎛ ∂u ⎞ 2 ∞ ∂u
π ∫0 ∂x
⇒ u c = −K ⎜ ⎟ + Ks sin sx dx
dt π ⎝ ∂x ⎠ x =0
⎡ ∂u ⎤
⎢⎣By assu min g ∂x → 0 as x → ∞ ⎥⎦
∞
d 2 2
⇒ uc = K μ + Ks [u sin sx ]∞0 − Ks 2 ∫ us cos sx dx
dt π π π0
2
=K μ − K s 2 u c if u→0 as x→∞
π
d 2
⇒ uc + K s2 uc = K μ …(2)
dt π
2 μ 2
uc = 2
(1 − e −Ks t )
πs
2μ ∞ cos sx − Ks 2 t
π ∫0 s 2
u(x, t) = (1 − e ) ds
2 μ
Put t = 0 ⇒ u c (s,0) = A + …(4)
π s2
from condition (i),
∞
u c (s, 0) = ∫ u ( x ,0) cos sx dx = 0 …(5)
0
2 μ
(4), (5) ⇒ A = −
π s2
37
2 μ − k 2st
∴ (3) ⇒ u c = (1 − e )
π s2
Finite Fourier Transform :-
Finite Fourier Since Transform :- Let f(x) denote a function that is sectionally continuous over
some finite interval (0, λ) of the variable x. The finite fourier sine transform of f(x) on the
interval is defined as
λ
sπx
fs(s) = ∫ f ( x ) sin dx where s is an integer
0 λ
Inversion formula for sine transform
2 ∞ sπx
f(x) = ∑
λ s=1
f s (s) sin
λ
for the interval (0, λ).
38
1 1
= [ − cos sπ + cos 0] = [1 − ( −1) 2 ]
s s
π π
⎛ sin sx ⎞
and fc(s) = ∫ cos sx dx = ⎜ ⎟ = 0 if s = 1, 2, 3,….
0 ⎝ s ⎠0
If s = 0, then
π
fc(s) = ∫1. dx = π
0
∂u ∂ 2 u
Find Finite FST and finite FCT of , , u(x, t) for 0 < x < λ, t > 0.
∂x ∂x 2
∂u
Solution :- By definition, finite FST of is
∂x
λ
∂u sπx
fs(s) = ∫ ∂x sin λ
dx
0
λ
sπx sπ λ sπx
= sin u ( x , t ) − ∫ u ( x , t ) cos dx
λ 0 λ0 λ
λ
∂u ⎤ − sπ
fs(s) = Fs ⎡
sπx
⇒
⎢⎣ ∂x ⎥⎦ = λ ∫ u(x, t) cos
λ
dx
0
⎡ ∂u ⎤ − sπ
⇒ fs(s) = Fs ⎢ ⎥ = Fc [u ( x , t )] …(1)
⎣ ∂x ⎦ λ
λ λ λ
⎡ ∂u ⎤ ∂u sπx sπx sπ sπx
Fc ⎢ ⎥ = ∫ cos dx = cos u ( x , t ) + ∫ u ( x , t ) sin dx
⎣ ∂x ⎦ 0 ∂x λ λ 0 0 λ λ
sπ
= Fs[u(x, t)] − [u(0, t) − u(λ, t) cos sπ] …(2)
λ
∂ 2u
To calculate the finite FST & finite FCT of ,
∂x 2
∂u
Replace u by in (1) and (2), we get
∂x
⎡ ∂ 2 u ⎤ − sπ ⎡ ∂u ⎤
Fs ⎢ 2 ⎥ = Fc ⎢ ⎥
⎣ ∂x ⎦ λ ⎣ ∂x ⎦
39
− sπ ⎡ sπ ⎤ sπ
=
λ ⎢⎣ λ
Fs ( u ) ⎥⎦ + λ {u (0, t ) − u (λ, t ) cos sπ}
− s2π2 sπ
= Fs [u ] + [u(0, t) − u(λ, t) cos sπ] …(3)
λ2
λ
⎡ ∂ 2 u ⎤ + sπ ⎡ ∂u ⎤
Similarly Fc ⎢ 2⎥
= Fs ⎢ ⎥ − [ux(0, t) − ux(λ, t) cos sπ]
⎣ ∂x ⎦ λ ⎣ ∂x ⎦
sπ ⎛ − sπ ⎞
= ⎜ ⎟ Fc[u(x, t)] − {ux(0, t) − ux(λ, t) cos sπ]
λ⎝ λ ⎠
− s2π2
= Fc[u] − {ux(0, t) −ux(λ, t) cos sπ} …(4)
λ2
Example:- Use finite Fourier transform to solve
∂u ∂ 2 u
= …(1)
∂t ∂x 2
u(0, t) = 0, u(4, t) = 0
u(x, 0) = 2x when 0 < x < 4, t > 0.
Solution :- (u)x = 0 is given, we apply FST.
Here λ = 4.
Taking Finite FST of (1), we get
4 4 2
∂u ⎛ sπx ⎞ ∂ u ⎛ sπx ⎞
∫ ∂t sin⎜⎝ 4 ⎟⎠dx = ∫ dx 2 sin⎜⎝ 4 ⎟⎠dx
0 0
d − s2π2 sπ
us = u s + (0, − 0)
dt 16 4
[Using equation (3) in previous article and given condition]
d s2π2
⇒ us + us = 0
dt 16
−s 2 π 2 t
⇒ u s (s, t ) = E e 16
…(∗)
40
4
⎛ sπx ⎞
u s (s,0) = ∫ 2 x sin ⎜ ⎟dx
0 ⎝ 4 ⎠
4
⎡ ⎛ sπx ⎞ 4 ⎤ 4 4 πsx
= 2 ⎢ x ⎜ − cos ⎟. ⎥ + 2. ∫ cos dx
⎣ ⎝ 4 ⎠ sπ ⎦ 0 sπ 0 4
4
16 8 ⎡ ⎛ sπx ⎞ 4 ⎤
= 2(−1) s+1
+ sin ⎜ ⎟.
s π sπ ⎢⎣ ⎝ 4 ⎠ sπ ⎥⎦ 0
16 8
= 2(−1)s+1. + (0)
s π sπ
32 32
= (−1)s+1 = − cos sπ
sπ sπ
from (∗), u s (s, 0) = A
− 32
∴ A= cos sπ
sπ
−s 2 π 2
32 t
Hence u s (s, t ) = − cos sπe 16
sπ
Taking Inverse Finite Fourier sine transform, we get
−s 2 π 2
2 ∞ ⎛ − 32 ⎞ t ⎛ sπx ⎞
u(x, t) = ∑ ⎜ cos sπ ⎟ e 16
sin ⎜ ⎟
4 s=1 ⎝ sπ ⎠ ⎝ 4 ⎠
−s 2 π 2 t
− 16 ∞ cos sπ ⎛ sπx ⎞
∴ u(x, t) = ∑ sin ⎜ ⎟e 16
π s=1 s ⎝ 4 ⎠
Example :- Solve the equation
∂u ∂ 2u
=2 2 …(1)
∂t ∂x
subject to conditions
(i) u(0, t) = 0
(ii) u(x, 0) = e−x
(iii) u(x, t) is bounded when x > 0, t > 0
Solution :- Since (u)x = 0 is given, so taking FST, we get
41
2 ∞ ∂u 2 ∞ ∂2u
π ∫0 ∂t π ∫0 ∂x 2
sin sx dx = 2 sin sx dx
2 d ∞ 2 ∞ ∂ 2u
π dt ∫0 π ∫0 ∂x 2
⇒ u sin sx dx = 2 sin sx dx
d 2∞ 2 ∞
∂ 2u
⇒ ∫
dt π 0
u sin sx dx = 2
π ∫ ∂x 2 sin sx dx
0
2 ⎡⎛ ∂u ⎤
∞ ∞
d ∂u ⎞
⇒ u s = 2 ⎢⎜ sin sx ⎟ − s ∫ cos sx dx ⎥
dt π ⎢⎣⎝ ∂x ⎠ 0 0 ∂x ⎥⎦
2 2 ∞ ∂u
=2 (0) − 2 s ∫ cos sx dx
π π 0 ∂x
⎡ ∂u ⎤
⎢⎣assu min g ∂x → 0 as x → ∞ ⎥⎦
d 2⎡ ∞
⎤
⎢− s(u ( x, t ) cos sx ) 0 − s ∫ sin sxu ( x, t )dx ⎥
∞
⇒ us = 2 2
dt π⎣ 0 ⎦
2⎡ 2∞ ⎤
=2 ⎢− s ∫ u ( x, t ) sin sxdx ⎥ [assuming u(x, t) →0 as x→∞]
π⎣ 0 ⎦
2∞
π ∫0
= −2s2 u ( x , t ) sin sx dx
d
⇒ u s = −2s 2 u s
dt
d
⇒ u s + 2s 2 u s = 0 …(2)
dt
A.E. is D + 2s2 = 0
2
Solution is u s (s, t ) = Ae−2s t
…(3)
To find A, taking FST of condition
(ii) u(x, 0) = e−x
Taking FST, we get
42
2 ∞ −x
π ∫0
u s (s,0) = e sin sx dx
2 s
= at t = 0 …(4)
π 1 + s2
⎡ e ax
Θ ∫ =
ax
⎢ u sin g e sin bx dx (a sin bx − b cos bx) taking a = −1, b = s
⎣ a 2
+ b 2
2 s ⎤
and then taking lim, we get ⎥
π 1 + s2 ⎦
Putting t = 0 in (3), we get
u s (s, 0) = A …(5)
from (4) and (5), we get
2⎛ s ⎞
A= ⎜ ⎟
π ⎝ 1 + s2 ⎠
Putting value of A in (3), we get
2 ⎛ s ⎞ −2s2 t
u s (s, t ) = ⎜ ⎟e
π ⎝1 + s2 ⎠
Taking inverse FST, we get
2 ∞ 2 ⎛ s ⎞ −2s2 t
π ∫0 π ⎝ 1 + s 2 ⎠
u(x, t) = ⎜ ⎟e sin sx ds
2∞ s 2
= ∫
π 01+ s 2
e −2s t sin sx ds which is required solution.
∂u ∂ 2 u
= …(1)
∂t ∂x 2
subject to conditions
(i) ux(0, t) = 0
⎧x , 0 < x ≤1
(ii) u(x, 0) = ⎨
⎩0 , x >1
43
(iii) u(x, t) is bounded
Solution :- Since (ux)x = 0 is given, so taking FCT of (1), we get
2 ∞ ∂u 2 ∞ ∂ 2u
π ∫0 ∂t π ∫0 ∂x 2
cos sx dx = cos sx dx
2 d ∞ 2 ∞ ∂ 2u
π dt ∫0 π ∫0 ∂x 2
⇒ u cos sx dx = cos sx dx
2 ⎡⎛ ∂u ∂u ⎤
∞ ∞
d ⎞
⇒ uc = ⎢⎜ cos sx ⎟ + s ∫ sin sx dx ⎥
dt π ⎣⎢⎝ ∂x ⎠0 0 ∂x ⎦⎥
∞
d 2
⇒ uc = (−s 2 ) ∫ cos sx u(x, t) dx
dt π 0
2∞
π ∫0
= (− s 2 ) u ( x , t ) cos sx dx
d
⇒ u c = −s 2 u c
dt
d
⇒ uc + s2 uc = 0 …(2)
dt
A.E. is D + s2 = 0
2
∴ solution is u c (s, t ) = Ae−s t
…(3)
2∞
π ∫0
u c (s,0) = u ( x ,0) cos sx dx
21
π ∫0
= x cos sx dx [Θ u(x, 0) = 0, x > 1]
2 ⎡ ⎡ sin sx ⎤ sin sx ⎤
1 1
s ⎥⎦ 0 ∫0 s
= ⎢⎢x − dx ⎥
π ⎢⎣ ⎣ ⎥⎦
44
2 ⎧⎪ sin s ⎡ cos sx ⎤ ⎫⎪
1
= ⎨ +⎢ 2 ⎥ ⎬
π ⎪⎩ s ⎣ s ⎦ 0 ⎪⎭
2 ⎡ sin s cos s − 1⎤
= + at t = 0 …(4)
π ⎢⎣ s s 2 ⎥⎦
Putting t = 0 in (3), we get
2 ⎡ sin s cos s − 1⎤
u c (s, 0) = A ⇒ A = +
π ⎢⎣ s s 2 ⎥⎦
from (3) and (4), we get
2∞
π ∫0
u(x, t) = u c (s, t ) cos sx ds
∂u ∂ 2 u
= , x > 0, y > 0 …(1)
∂t ∂x 2
Subject to condition
(i) u(0, t) = 0
⎧1 , 0 < x <1
(ii) u(x, 0) = ⎨ when t = 0
⎩0 , x ≥1
(iii) u(x, t) is bounded.
Solution :- Since (u)x = 0 is given, so taking FST of (1), we get
∞
2 ∂u 2 ∞ ∂ 2u
π ∫ ∂t sin sx dx = π ∫0 ∂x 2
sin sx dx
0
2 d ∞ 2 ∞ ∂ 2u
π dt ∫0 π ∫0 ∂x 2
⇒ u sin sx dx = sin sx dx
45
∞
d 2 ∞ 2 ⎡ ∂u ⎤ 2 ∞ ∂u
⇒
dt π ∫0
u sin sx dx =
π ⎢⎣sin sx ∂x ⎥⎦ − ∫
π 0
s cos sx dx
∂x
0
d 2 ∞ ∂u ⎡ ∂u ⎤
π ∫0 ∂x
⇒ u s = ( 0) − s cos sx dx ⎢⎣assuming ∂x → 0 as x → ∞ ⎥⎦
dt
∞
2
= −s [cos sx.u ( x, t )]0∞ + s 2 ∫ (−s sin sx )u ( x, t )dx
π π0
∞
d 2 2 2
π ∫0
⇒ u s = s u (0, t ) − s u sin sx dx
dt π
= 0 − s2 u s [Θ u(0, t) = 0 given]
d
⇒ us + s2 us = 0 …(2)
dt
A.E. is D + s2 = 0
2
∴ Solution is u s (s, t ) = Ae−s t
…(3)
2∞
π ∫0
u s (s,0) = u ( x ,0) sin sx dx
21
π ∫0
= (1) sin sx dx [Θ u(x, 0) = 0 for x ≥ 1]
1
2 ⎡ − cos sx ⎤
=
π ⎢⎣ s ⎥⎦
0
2 ⎡ cos s 1 ⎤ 2 ⎛ 1 − cos s ⎞
⇒ u s (s,0) = − + ⎥= ⎜ ⎟ …(4)
π ⎢⎣ s s⎦ π⎝ s ⎠
From (3), we have u s (s,0) = A …(5)
2 ⎛ 1 − cos s ⎞
A= ⎜ ⎟
π⎝ s ⎠
2 ⎛ 1 − cos s ⎞ −s2t
∴ (3) ⇒ u s (s, t ) = ⎜ ⎟e
π⎝ s ⎠
Taking inverse FST, we get
46
2 ∞ 2 ⎛ 1 − cos s ⎞ −s2t
π ∫0 π ⎝
u(x, t) = ⎜ ⎟e sin sx ds
s ⎠
2 ∞ ⎛ 1 − cos s ⎞ −s 2 t
⇒ u(x, t) = ∫ ⎜ ⎟ sin sx e ds
π 0⎝ s ⎠
which is required solution.
47
LESSON 3 CURVILINEAR CO-ORDINATES
Transformation of coordinates
Let the rectangular coordinates (x, y, z) of any point be expressed as functions of (u1, u2,
u3) so that
x = x (u1 , u 2 , u 3 )⎫
⎪
y = y(u 1 , u 2 , u 3 ) ⎬ …(1)
z = z(u 1 , u 2 , u 3 ⎪⎭
Suppose (1) can be solved for u1, u2, u3 in terms of x, y, z i.e.
u 1 = u 1 ( x , y, z ) ⎫
⎪
u 2 = u 2 ( x , y, z ) ⎬ …(2)
u 3 = u 3 ( x , y, z) ⎪⎭
Here correspondence between (x, y, z) and (u1, u2, u3) is unique i.e. if to each point P(x, y, z) of
some region R, there corresponds one & only one triad (u1, u2, u3), then u1, u2, u3) are said to be
curvilinear coordinates of the point P.
The set of equations (1) & (2) define a transformation of co-ordinates.
z u3 curve
P(x,y,z)
(u1,u2,u3)
u1 curve u2 curve
0 y
Co-ordinate surfaces
x and curves:-
The surfaces
u 1 = c 1, u 2 = c 2, u 3 = c 3
where c1, c2, c3 are constants i.e. surfaces whose equations are
u1 = u1(x, y, z) = c1
u2 = u2(x, y, z) = c2
u3 = u3(x, y, z) = c3
48
are called co-ordinate surfaces and each pair of these surfaces intersect in curves called co-
ordinate curves or lines.
So u1 = c1 and u2 = c2 gives u3−curve
Similarly u2 = c2 and u3 = c3 gives u1 − curve
and u1 = c1 and u3 = c3 gives u2− curve
So if u2 = c2 and u3 = c3 and if u1 is the only variables then point P describe a curve known as u1
curve which is a function of u1.
z
u3−curve
u2= c2 u1= c1
u3 = c3
u2−curve
u1−curve
0 y
x
Orthgonal Curvilinear Co-ordinates
If the three co-ordinate surfaces intersect at right angles (orthogonal), the curvilinear co-
ordinate system is called orthogonal i.e. co-ordinates (u1, u2, u3) are said to be orthogonal
curvilinear co-ordinates.
Then u1, u2, u3 co-ordinate curves of a curvilinear system are analogous to the x, y, z –
coordinates axes of rectangular system.
Unit vectors in curvilinear system
ρ
Let r = x î + y ĵ + z k̂
be position vector of P, then (1) can be written as
ρ ρ
r = r (u1, u2, u3)
ρ
∂r
A tangent vector to the u1-curve at P(for which u2 & u3 are constant) is
∂u 1
Then, A unit tangent vector in this direction is
49
ρ
∂ r / ∂u 1
ê1 = ρ
∂r
∂u 1
ρ ρ
∂r ∂r
or = ê1
∂u 1 ∂u 1
ρ ρ
∂r ∂r
⇒ = h 1 ê1 where h 1 =
∂u 1 ∂u 1
Similarly if ê 2 and ê 3 are unit tangent vectors to the u2 − & u3− curves at P respectively, then
ρ ρ
∂r ∂r
= h 2 ê 2 , = h 3 ê 3
∂u 2 ∂u 3
ρ ρ
∂r ∂r
where h2 = , h3 =
∂u 2 ∂u 3
The quantities h1, h2, h3 are called scale factors.
Also, condition for the orthogonality of co-ordinate surfaces are
ρ ρ ρ ρ ρ ρ
∂r ∂r ∂r ∂r ∂r ∂r
. = 0, . = 0, . =0
∂u 1 ∂u 2 ∂u 2 ∂u 3 ∂u 3 ∂u 1
Also the vectors ∇u1, ∇u2, ∇u3 at P are directed along the normal to the co-ordinate surfaces
u1 = c1, u2 = c2, u3 = c3 respectively.
So unit-vectors in these directions are given by
∇u 1 ∇u 2 ∇u 3
Ê1 = , Ê 2 = , Ê 3 =
| ∇u 1 | | ∇u 2 | | ∇u 3 |
z u3−curve
ê 3
Ê 3 Ê 2
u1−curve ê 2
50
x
Thus, at each point P of a curvilinear system, there are, in general, two sets of unit vectors
ê1 , ê 2 , ê 3 tangent to the coordinate curves and Ê1 . Ê 2 , Ê 3 normal to the co-ordinate surfaces.
The sets become identical if & only if the curvilinear coordinate system is orthogonal.
Arc length and Volume element
ρ ρ
From r = r (u 1 , u 2 , u 3 )
ρ ρ ρ
ρ ∂r ∂r ∂r
⇒ dr = du 1 + du 2 + du 3 …(1)
∂u 1 ∂u 2 ∂u 3
= h1 du1 ê1 + h 2 du 2 ê 2 + h 3 du 3 ê 3
51
(ds)2 = h 1 (du1 ) + h 2 (du 2 ) + h 3 (du 3 )
2 2 2 2 2 2
⇒
ρ
∂r 1
where h1 = =
∂u 1 | ∇u 1 |
ρ
∂r 1
h2 = =
∂u 2 | ∇u 2 |
ρ
∂r 1
h3 = =
∂u 3 | ∇u 3 |
ρ ρ
Now if r = r (x, y, z)
∂u 1 ∂u 1 ∂u ∂u
Then ρ = î + 1 ĵ + 1 k̂ = ∇u1
∂r ∂x ∂y ∂z
ρ
∂r 1
⇒ =
∂u 1 ∇u 1 u3−curve
h3du3 u2−curve
h1du1 h2du2
P
u1−curve
52
dV = |(h1 du1 ê1 ).(h 2 du 2 ê 2 )(h 3 du 3 ê 3 ) | [Θ volume = | â.(b̂ × ĉ) |
∂φ ∂φ ∂u 1 ∂φ ∂u 2 ∂φ ∂u 3
= + + …(2)
∂y ∂u 1 ∂y ∂u 2 ∂y ∂u 3 ∂y
∂φ ∂φ ∂u 1 ∂φ ∂u 2 ∂φ ∂u 3
= + + …(3)
∂z ∂u 1 ∂z ∂u 2 ∂z ∂u 3 ∂z
∂φ ∂φ ∂φ
∇φ = grad φ = ∇u 1 + ∇u 2 + ∂u 3
∂u 1 ∂u 2 ∂u 3
But ê1 = h 1∇u 1 , ê 2 = h 2 ∇u 2 , ê 3 = h 3 ∇u 3
ê1 ∂φ ê 2 ∂φ ê 3 ∂φ
⇒ grad φ = ∇φ = + +
h 1 ∂u 1 h 2 ∂u 2 h 3 ∂u 3
Example:- If (u1, u2, u3) are orthogonal coordinate then prove that
(i) |∇up| = hp−1, p = 1, 2, 3
(ii) ê p = Ê p
53
| ê1 | 1
|∇u1| = = = h 1−1
| h1 | h1
Similarly φ = u2, then
1 ∂u 2 ê
∇u 2 = ê 2 = 2
h 2 ∂u 2 h2
| ê 2 | 1
⇒ |∇u2| = = = h −21
| h2 | h2
Similarly if φ = u3, then |∇u3| = h3−1
∇u p
(ii) By definition, Ê p =
| ∇u p |
⇒ Ê p = h p ∇u p = ê p
Results :-
= φ ∇. f̂ + f̂ . ∇φ
ρ ρ ρ ρ
II. div ( f̂ × g ) = ∇. ( f̂ × g ) = curl f̂ . g − curl g . f̂
ê 2 = h2 ∇u2
ê 3 = h3 ∇u3
ρ
f = f1 h2 h3(∇u2 × ∇u3) + f2 h3 h1 (∇u3 × ∇u1) + f3 h1 h2(∇u1 × ∇u2)
54
ρ
∴ dif f = ∇ . f = ∇.[f1h2 h3(∇u2 × ∇u3)]
+ ∇ . [f2 h3 h1 (∇u3 × ∇u1)]
+ ∇. [f3 h1 h2 (∇u1 × ∇u2)] …(∗)
Taking first art,
∇ . [f1 h2 h3(∇u2 × ∇u3)] = f1 h2 h3 ∇ . (∇u2 × ∇u3) + (∇u2 × ∇u3). ∇(f1h2h3)
ρ ρ ρ
[By using ∇. (φ f ) = φ ∇. f + f . ∇φ]
⇒ ∇.[f1 h2 h3(∇u2 × ∇u3)] = f1 h2 h3 ∇.(∇u2 × ∇u3) + (∇u2 × ∇u3).
∇(f1 h2h3) …(1)
⇒ ∇.[(∇u2 × ∇u3)] = curl ∇u2 . ∇u3 − curl ∇u3 . ∇u2
= (curl grad u2). (∇u3) − (curl grad u3). ∇u2
=0−0=0
∴ from (1), we get
∇ . [f1 h2 h3 (∇u2 × ∇u3)] = (∇u2 × ∇u3). ∇(f1 h2 h3)
⎡ ∂ ∂
= (∇u2× ∇u3). ⎢ (f1 h2 h3) ∇u1 + (f1 h2 h3) ∇u2
⎣ ∂u1 ∂u 2
∂ ⎤
+ (f1h 2 h 3 )∇u 3 ⎥
∂u 3 ⎦
ê 2 × ê 3 ⎡ ∂ ê ∂ ê ∂ ê ⎤
= ⎢ ( f1 h 2 h 3 ) 1 + ( f1 h 2 h 3 ) 2 + ( f1 h 2 h 3 ) 3 ⎥
h 2 h 3 ⎣ ∂u1 h1 ∂u 2 h 2 ∂u 3 h3 ⎦
⎡ ê1 ê 2 ê ⎤
⎢Θ ∇u 1 = , ∇u 2 = , ∇u 3 = 3 ⎥
⎣ h1 h2 h3 ⎦
ê1 ⎡ ∂ ê ∂ ê ∂ ê ⎤
= .⎢ (f 1 h 2 h 3 ) 1 + ( f1 h 2 h 3 ) 2 + ( f1 h 2 h 3 ) 3 ⎥
h 2 h 3 ⎣ ∂u1 h1 ∂u 2 h 2 ∂u 3 h3 ⎦
1 ∂
= ( f1 h 2 h 3 )
h 1h 2 h 3 ∂u 1
1 ∂
⇒ ∇.[f1 h2 h3(∇u2 × ∇u3)] = . (f1 h2 h3)
h 1h 2 h 3 ∂u 1
1 ∂
Similarly ∇.[f2 h3 h1(∇u3 × Δu1)] = (f2 h3 h1)
h 1h 2 h 3 ∂u 2
55
1 ∂
and ∇.[f3 h1 h2(∇u1 × Δu2)] = (f3 h1 h2)
h 1h 2 h 3 ∂u 3
ê 2 ∂ ê ∂
⇒ ∇×(h1f1 ∇u1) = ( f1 h 1 ) − 3 . (f1h 1 )
h1h 3 ∂u 3 h 2 h1 ∂u 2
ê 3 ∂ − ê1 ∂
Similarly ∇×(h1f2 ∇u2) = (f 2 h 2 ) (f 2 h 2 )
h 2 h1 ∂u1 h 3 h 2 ∂u 3
ê1 ∂ ê ∂
and ∇× (h3 f3 ∇u3) = (f 3 h 3 ) − 2 (f 3 h 3 )
h 3 h 2 ∂u 2 h 1h 3 ∂u 1
∴ from (1),
56
ρ ê ⎡ ∂ ∂ ⎤
Curl f = 1 ⎢ (f 3 h 3 ) − (f 2 h 2 ) ⎥
h 2 h 3 ⎣ ∂u 2 ∂u 3 ⎦
ê 2 ⎡ ∂ ∂ ⎤
+ ⎢ ( f1 h 1 ) − (f 3 h 3 ) ⎥
h 3 h1 ⎣ ∂u 3 ∂u1 ⎦
ê 3 ⎡ ∂ ∂ ⎤
+ ⎢ (f 2 h 2 ) − (f1 h 1 ) ⎥
h1h 2 ⎣ ∂u1 ∂u 2 ⎦
h 1ê1 h 2 ê 2 h 3 ê 3
1 ∂ ∂ ∂
=
h 1h 2 h 3 ∂u 1 ∂u 2 ∂u 3
h 1f 1 h 2f 2 h 3f 3
⎡ 1 ∂φ 1 ∂φ 1 ∂φ ⎤
= ∇. ⎢ ê1 + ê 2 + ê 3 ⎥
⎣ h1 ∂u1 h 2 ∂u 2 h 3 ∂u 3 ⎦
We know that
ρ 1 ⎡ ∂ ∂ ∂ ⎤
∇. f = ⎢ ( f1 h 2 h 3 ) + (f 2 h 3 h 1 ) + (f 3 h 1 h 2 ) ⎥
h 2 h 1h 3 ⎣ ∂u1 ∂u 2 ∂u 3 ⎦
1 ∂φ
Here put f1 = ,
h 1 ∂u1
1 ∂φ 1 ∂φ
f2 = , f3 =
h 2 ∂u 2 h 3 ∂u 3
Therefore
1 ⎡ ∂ ⎛ h 2 h 3 ∂φ ⎞ ∂ ⎛ h 3 h 1 ∂φ ⎞ ∂ ⎛ h 1h 2 ∂φ ⎞⎤
∇ 2φ = ⎢ ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟⎥
h 1h 2 h 3 ⎣ ∂u 1 ⎝ h 1 ∂u 1 ⎠ ∂u 2 ⎝ h 2 ∂u 2 ⎠ ∂u 3 ⎝ h 3 ∂u 3 ⎠⎦
57
CYLINDRICAL POLAR COORDINATES (r, θ, z)
Let P is a point having Cartesian co-ordinates (x, y, z)
OM = x, MN = y(= OQ), PN = z z
P(x,y,z)
0 y z Q
N y
x θ r
r M
y
N
x
θ
M x 0
x = OM = ON cos θ
⇒ x = r cos θ …(1)
& y = MN = N sin θ
⇒ y = r sion θ …(2)
and z=z …(3)
so we have,
r= x 2 + y2
y
and θ = tan−1
x
Determine the transformation from cylindrical to rectangular coordinates :- Operating equation
(1)2 + (2)2, we get
r2 = x 2 + y 2 ⇒ r = x 2 + y2
y ⎛y⎞
(2)/(1) ⇒ = tanθ ⇒ θ = tan −1 ⎜ ⎟
x ⎝x⎠
and z = z.
(ds)2 = (dx)2 + (dy)2 (dz)2 …(4)
Now dx = cos θ dr − r sin θ dθ
dy = sinθ dr + r cos θ dθ
58
dz = dz
∴ (4) ⇒
∴ (ds)2 = (dr cos θ − r sin θ dθ)2 + (sinθ dr + r cos θdθ2 + (dz)2
= (dr)2 (sin2θ + cos2θ) + (r dθ)2 (sin2θ + cos2θ) + (dz)2
⇒ (ds)2 = (dr)2 + (rdθ)2 + (dz)2
We get,
h1 = 1, h2 = r, h3 = 1, u1 = r, u2 = θ, u3 = z
Take ê1 = ê r , ê 2 = ê θ , ê 3 = ê z
1 ⎡ ∂ ⎛ ∂Φ ⎞ ∂ ⎛ 1 ∂Φ ⎞ ∂ ⎛ ∂Φ ⎞⎤
∇ 2Φ = ⎜r ⎟+ ⎜ ⎟ + ⎜r ⎟
r ⎢⎣ ∂r ⎝ ∂r ⎠ ∂θ ⎝ r ∂θ ⎠ ∂z ⎝ ∂z ⎠⎥⎦
∂ 2 Φ 1 ∂Φ 1 ∂ 2 Φ ∂ 2 Φ
= + + +
∂r 2 r ∂r r 2 ∂θ 2 ∂z 2
SPHERICAL POLAR COORDINATES (r, θ, φ)
z
(r,θ,φ)
P(x,y,z)
θ
r O
0 y r
φ z
x θ
N P
M
x y N
59
OM = x, MN = y, PN = z, OP = r z = PN = r cos θ ON = r sinθ
In Δ OMN, x = OM = ON cos φ
O
x φ
M y N
⇒ x = r sinθ cosφ
y = MN = ON sinφ ⇒ y = r sin θ sin φ
Now (ds)2 = (dx)2 + (dy)2 + (dz)2 …(1)
dx = sinθ cosφ dr + r cosθ cos φ dθ − r sin θ sin φ dφ
dy = sinθ sinφ dr + r cosθ sinφ dθ + r sinθ cosφ dφ
dz = cosθ dr − r sin θ dθ …(2)
Put the value of (2) in (1) and collecting the coefficients of (dr)2, (dθ)2, (dφ)2, we get
(ds)2 = (dr)2 (1) + (dθ)2 (r2) + (dφ)2 (r2 sin2θ)
Comparing it with,
(ds)2 = h12 (du1 ) 2 + h 22 (du 2 ) 2 + h 32 (du 3 ) 2
we get
h1 = 1, h2 = r, h3 = r sin θ
u1 = r, u2 = θ, u3 = φ
ê1 = ê r , ê 2 = ê θ , ê 3 = ê φ
∂Φ ê θ ∂Φ ê φ ∂Φ
So grad Φ = ê r + +
∂r r ∂θ r sin θ ∂φ
ρ 1 ⎡∂ 2 ∂ ∂ ⎤
div f = 2 ⎢ ∂r (f1r sin θ) + ∂θ (f 2 r sin θ + ∂φ (f 3 r )⎥
r sin θ ⎣ ⎦
ρ 1 ⎡ ∂ ⎛ r 2 sin θ ∂Φ ⎞⎤
Curl f = ⎢ ⎜ ⎟⎥
r 2 sin θ ⎢⎣ ∂r ⎜⎝ (1) ∂r ⎟⎠⎥⎦
60
ê r rê θ r sin θ ê φ
ρ 1 ρ 1 ∂ ∂ ∂
Curl f = f=
2
r sin θ r ∂r ∂θ ∂φ
f1 f2r f 3 r sin θ
1 ⎡∂ ⎛ 2 ∂Φ ⎞ ∂ ⎛ ∂Φ ⎞ ∂ ⎛ 1 ∂Φ ⎞⎤
∇ 2Φ = ⎢ ⎜ r sin θ ⎟ + ⎜ sin θ ⎟+ ⎜ ⎟⎥
r sin θ ⎣ ∂r ⎝
2
∂r ⎠ ∂θ ⎝ ∂θ ⎠ ∂φ ⎜⎝ sin θ ∂φ ⎟⎠⎦
∂ 2 Φ 2 ∂Φ 1 ∂ 2 Φ cos θ ∂Φ 1 ∂ 2Φ
= + + + +
∂r 2 r ∂r r 2 ∂θ 2 r 2 sin θ ∂θ r 2 sin 2 θ ∂φ 2
Example:- Prove that cylindrical coordinate system is orthogonal.
Solution :- The position vector of any point in cylindrical coordinate is
ρ
r = x î + y ĵ + z k̂
ρ
∴ r = r cos θ î + r sin θ ĵ + z k̂
ρ ρ ∂x ∂y ∂z
r = r (s) , , are unit vector
∂s ∂s ∂s
The tangent vector to the r, θ and z curves are given by
ρ ρ ρ
∂r ∂r ∂r
, , respectively.
∂r ∂θ ∂z
ρ
∂r
where = cos θ î + sin θ ĵ
∂r
ρ
∂r
= − r sin θ î + r cos θ ĵ
∂θ
ρ
∂r
= k̂
∂z
Therefore the unit-vector in these tangent directions are
ρ
∂ r ∂r
ê1 = ê r =
∂r ∂r
cos θ î + sin θ ĵ
⇒ ê r = = cos θ î + sin θ ĵ
2 2
cos θ + sin θ
ρ
∂ r / ∂θ − r sin θ î + r cos θ ĵ
ê 2 = ê θ = ρ =
∂r r 2 sin 2 θ + r 2 cos 2 θ
∂θ
61
− r sin θ î + r cos θ ˆj
⇒ ê θ = = − sin θ î + cos θ ˆj
r
ρ
∂ r / ∂z
ê 3 = ê z = ρ = k̂
| ∂ r / ∂z |
= sin2θ + cos2θ = 1
ê z .ê z = k̂.k̂ = 1
This shows that ê r , ê θ , ê z are mutually ⊥r and therefore the coordinate system is orthogonal.
Example:- (a) Find the unit vectors ê r , ê θ , ê φ of a spherical co-ordinate system in terms of
î , ĵ, k̂ .
Solution :- The position vector of any point in spherical coordinate is
ρ
R = x î + y ˆj + z k̂
62
ρ
∂R / ∂θ r cos θ cos φ î + r cos θ sin φ ĵ − r sin θk̂
So ê θ = ρ =
| ∂R / ∂θ | r 2 cos 2 θ(cos 2 φ + sin 2 φ) + r 2 sin 2 θ
63
= sin2θ(cos2φ + sin2φ) + cos2θ
⇒ ê r .ê r = 1
= sin2φ + cos2φ = 1
⇒ ê φ .ê φ = 1
This shows that ê r , ê θ , ê φ are mutually and therefore the coordinate system is orthogonal.
ρ
Example:- Represent the vector A = 2 y î − z ˆj + 3x k̂ …(1)
in spherical coordinates
Solution :- Here x = r sinθ cosφ
y = r sinθ sinφ
z = r cosθ …(2)
and ê r = sinθ cosφ î + sinθ sinφ ˆj + cosθ k̂
64
+ 3 r sinθ cosφ ( ê r cos θ − ê θ sin θ)
ρ
⇒ A = ê r (2r sin2θ cos2φ − r sinθ cosθ sinφ + 3 r sinθ cosφ cosθ)
+ êθ (2r sinθ cosθ sinφ cosφ − r cos2θ sinφ − 3r sin2θ cosφ)
d
∴ ê r = (− sin θ)θ&î + (cos θ)θ&ˆj
dt
d
⇒ ê r = (−sinθ î + cosθ ˆj ) θ&= θ&ê θ [using (2)]
dt
d
Also ê θ = (− cos θ)θ&î + (− sin θ) θ&ĵ
dt
= −(cos θ î + sin θ ĵ)θ&
d
⇒ ê θ = −θ&ê r [using (1)]
dt
ρ ρ
Example:- Express the velocity v and acceleration a of a particle in cylindrical coordinates.
Solution :- Position vector of a particle P in rectangular coordinates
ρ
r = x î + y ˆj + z k̂
ρ
vector r in cylindrical coordinate system is
ρ
r = r cosθ(cosθ ê r − sin θ ê θ )
65
+ r sin θ(sin θ ê r + cosθ ê θ ) + z ê z
ρ
⇒ r = r cos 2 θ ê r − r cosθ sinθ ê θ + r sin 2 θ ê r + r sin θ cos θ ê θ + z ê z
ρ
⇒ r = r ê r + z ê z
ρ
ρ dr
Then velocity v =
dt
ρ dr d dz d
⇒ v = ê r + r ê r + ê z + z ê z
dt dt dt dt
ρ ⎡ d ⎤
v = r&ê r + r θ&ê θ + z&ê z ⎢⎣Θ dt ê z = 0⎥⎦ …(1)
ρ ρ
ρ dv d 2 r
Differentiating (1) again, we obtain acceleration a = =
dt dt 2
ρ d
⇒ a = (r&ê r + r θ&ê θ + z ê z )
dt
d d d
r&ê r + r& ê r + r&θ&ê r + r &
=& θ&ê θ + r θ& ê θ &
z&ê z + z& ê z
dt dt dt
d d d
using ê r = θ&êθ , êθ = −θ&ê r , ê z = 0,
dt dt dt
we get
ρ
r&ê r + r&θ&ê θ + r&θ&ê θ + r&&
a =& θ&ê θ + r &
θ&ê θ + rθ&( −θ&ê r ) + &
z&ê z
ρ
⇒ r&− r θ&2 )ê r + (r &
a = (& θ&+ 2r&θ&)ê θ + &
z&ê z
66
= θ&(cos θ cos φ î + cos θ sin φ ĵ − sin θ k̂ )
= θ&(sin θ cos φ î + inθ sin φ ĵ + cos θ k̂ ) + cos θ φ&( − sin φ î + cos φ ĵ)
d
⇒ ê θ = −θ&ê r + cos θ ê φ φ&
dt
d
Also from (4) , ê φ = − cos φ φ&î − sin φ φ&− sin φ φ&ĵ …(5)
dt
Taking R.H.S of equation (*)
R.H.S = − sin θ φ&ê r − cos gq φ&ê θ
⇒ L.H.S. = R.H.S.
d
So ê φ = − sin θ φ&ê r − cos θ φ&ê θ
dt
ρ ρ
Example:- Express the velocity v and acceleration a of a particle in spherical coordinates.
Solution : Position vector of a particle P in rectangular coordinates
67
ρ ρ
r = x î + y ĵ + z k̂ vector r in spherical coordinate system is
ρ
r = r cos φ sinθ ( ê r sin θ cos φ + ê θ cos θ cos φ − ê φ sin φ)
68
+ r cos θ φ&ê φ θ&+ r sin θ φ
&
&ê φ
+ ê θ [2r&θ&+ r &
θ&− r sin θ cos θ φ&2 ]
⎡1 d 2 & ⎤
+ ê θ ⎢ (r θ) − r sin θ cos θ φ&2 ⎥
⎣ r dt ⎦
⎡ 1 d 2 2 &⎤
+ ê φ ⎢ (r sn θ φ)⎥ …(2)
⎣ r sin θ dt ⎦
⎡ 1 d 2 & 1 2& & & & & & 1 d 2
⎢⎣Θ r dt (r θ) = r (r θ + 2r r&θ) = r θ + 2r&θ and r sin θ dt (r sin θ φ&)
2
1
= &+ 2r r&sin 2 θ φ&+ φ&r 2 2 sin θ(cos θ)θ&]
&
[r 2 sin 2 θ φ
r sin θ
⎤
&+ 2 r&φ&sin θ + 2r θ&φ&cos θ ⎥
&
= r sin θ φ
⎦
So (2) is required expression for acceleration of a particle in spherical coordinate.
d
Example : ê φ = ê φ& = − sin θ φ&ê r − cos θ φ&ê θ
dt
Solution :- Now ê φ = − sin φ î + cos φ ˆj
d
⇒ ê φ = − cos φ φ&î − sin φ φ&ˆj …(5)
dt
Put values of î and ˆj in(5), we get
d
ê φ = − cos φ φ&(ê r sin θ cos φ + ê φ cos θ cos φ − ê φ sin φ)
dt
− sin φ φ&(ê r sin θ sin φ + ê θ . cos θ sin φ + ê φ cos φ)
69
= − sinθ φ&ê r − cos θ φ&ê θ . Hence Proved.
ρ
Example:- Represent the vector A = z î − 2 x ˆj + y k̂ in cylindrical coordinates.
Solution :- In cylindrical coordinates,
x = r cos θ, y = r sinθ, z = z …(1)
and ê r = cos θ î + sin θ ĵ
ê z = k̂ …(3)
The position vector of a particle P in rectangular coordinates is
ρ
A = z î + ( −2 x ) ĵ + y k̂
ρ
⇒ A = z î − 2r cos θ ˆj + r sin θ k̂ …(4)
Put the value of î , ĵ, k̂ from (3), (5), (6) in equation (4), we get
ρ
A = z(ê r cos θ − ê θ sin θ) − 2r cos θ(ê r sin θ + ê θ cos θ) + r sin θ ê z
where
Ar = z cos θ − 2r cosθ sinθ
Aθ = −z sinθ −2r cosθ sinθ
Az = r sinθ
70
ρ
Equation (1) is required vector A in cylindrical coordinates.
ρ
Example:- Represent the vector A = z î − 2 x ˆj + y k̂ in spherical coordinates (r, θ, φ)
Solution :- In spherical coordinates,
x = r sinθ cosφ
y = r sinθ sinφ
z = r cosθ …(2)
and
ê r = sinθ cosφ î + sinθ sinφ ĵ + cos θ k̂
71
sin θ cos φ ê r cos θ
D2 = cos θ cos φ ê θ − sin θ
− sin φ ê φ 0
= ê r cosθ(sin2φ + cos2φ) +
⇒ D3 = ê r cosθ − sinθ ê θ
D1
Then î = = ê r sinθ cosφ + ê θ cosθ cosφ − ê φ sinφ
D
D2
ĵ = = ê r sin θ sinφ + ê θ cosθ sinφ + cos φ ê φ
D
D3
k̂ = = ê r cosθ − ê θ sinθ
D
from (1), (2) and (4), we get
ρ
A ( r cos θ )(ê r sin θ cos φ + ê θ cosθ cosφ − ê φ sinφ)
72
ρ
⇒ A = r ê r (cosθ sinθ cosφ − 2 sin2θ cosφ sinφ + sinθ cosθ sinφ)
73
h1 = 1, h2 = r1, h3 = r sin θ
u1 = r, u2 = θ, u3 = φ
Example : - Find expression for the elements of area in orthogonal curvilinear coordinates.
Solution: -
u3 h3du3 ê 3
u2
P
h2du2 ê 2
h1du1 ê1
u1
Example:- Find the volume element dv in cylenderical & spherical coordinate system.
Solution: - We know that volume element dV in orthogonal curvilinear coordinates is
dV = h1 h2 h3 du1du2 du3 …(1)
In cylindrical coordinates (r, θ, φ)
h1 = 1, h2 = r , h3 = 1
u1 = r, u2 = θ, u3 = z
74
So (1) ⇒ dV = r dr dθ dz
In spherical coordinates (r, θ , φ)
h1 = 1, h2 = r, h3 = r sin θ
u1 = r, u2 = θ, u3 = φ
∴ So (1) ⇒ dV r2 sin2 dr dθ dφ
ρ ρ ρ
∂r ∂r ∂r
Example:- If u1, u2, u3 are general coordinates, show that , , & ∇u1,∇u2,∇u3 are
∂u 1 ∂u 2 ∂u 3
⎛ ∂φ ∂φ ˆ ∂φ ⎞
= ⎜⎜ î + j+ k̂ ⎟⎟ . (d x î + dy ˆj + dz k̂ )
⎝ ∂x ∂y ∂z ⎠
ρ
⇒ dφ = ∇φ.d r
Replacing φ with ∇ up, we get
ρ
dup = ∇up . d r …(1)
ρ
⇒ du1 = ∇u1 . d r
ρ
du2 = ∇u2 . d r
ρ
du3 = ∇u3 . d r
ρ ρ
Now r = r (u1, u2, u3)
ρ ρ ρ
ρ ∂r ∂r ∂r
⇒ dr = du 1 + du 2 + du 3 …(2)
∂u 1 ∂u 2 ∂u 3
75
Similarly taking dot product of (2) with ∇ u2 and ∇ u3, we get
ρ ρ ρ
∂r ∂r ∂r
∇ u 2, = 1, ∇u 2 . = 0 ∇u 2 . =0
∂u 2 ∂u 1 ∂u 3
ρ
∂r
∇ u 3. = 1,
∂u 3
So we get the required result i.e.
ρ
∂ r ⎧1 if p=q
∇ u p. =⎨
∂u q ⎩0 if p≠q
Example: - Prove that ê1 = h2 h3 ∇ u2 × ∇ u3 with similar equation for ê 2 and ê 3 where u1, u2, u3
ê 2 × ê 3 ê
Then ∇u2 × ∇u3 = = 1
h 2h 3 h 2h3
ê 3 kê1 ê
Similarly ∇u3 ×∇u1 = = 2
h 3 h1 h 3 h1
⇒ ê 2 = h 3 h 1∇u 3 × ∇u 1
ê1 × ê 2 ê
and ∇u1 × ∇u2 = = 3
h 1h 2 h 1h 2
Example: - If u1, u2, u3 are orthogonal curvilinear coordinates, show that the jacobian of x, y, z
w.r.t. u1, u2, u3 is
76
∂x ∂y ∂z
∂u 1 ∂u 1 ∂u 1
⎛ x , y, z ⎞ ∂ ( x , y, z ) ∂x ∂y ∂z
J ⎜⎜ ⎟⎟ = =
⎝ u 1u 2 u 3 ⎠ ∂ (u 1 , u 2 , u 3 ) ∂u 2 ∂u 2 ∂u 2
∂x ∂y ∂z
∂u 3 ∂u 3 ∂u 3
= h 1h 2h 3
ρ
Proof: - we know that A = A1 î + A2 ˆj +A3 k̂
ρ ρ
B = (B1 , B 2 , B3 ), C = C(C1 , C 2 , C3 )
Then
î ĵ k̂
ρ ρ ρ ρ
A.(B × C) = A. B1 B 2 B3
C1 C 2 C3
A1 A2 A3
ρ ρ ρ
⇒ A.(B × C) = B1 B2 B3
C1 C2 C3
Therefore
∂x ∂y ∂z
∂u1 ∂u1 ∂u1
∂x ∂y ∂z ⎛ ∂x ∂y ∂z ⎞ ⎡⎛ ∂x ∂y ∂z ⎞
= ⎜⎜ î + ŷ + k̂ ⎟⎟.⎢⎜⎜ î + ĵ + k̂ ⎟⎟
∂u 2 ∂u 2 ∂u 2 ⎝ ∂u1 ∂u1 ∂u1 ⎠ ⎣⎝ ∂u 2 ∂u 2 ∂u 2 ⎠
∂x ∂y ∂z
∂u 3 ∂u 3 ∂u 3
⎛ ∂x ∂y ˆ ∂z ⎞⎤
× ⎜⎜ î + j+ k̂ ⎟⎟⎥ ]
⎝ ∂u 3 ∂u 3 ∂u 3 ⎠⎦
ρ ρ ρ
⎛ ∂r ⎞ ⎛ ∂r ∂r ⎞
= ⎜⎜ ⎟⎟.⎜⎜ × ⎟⎟
⎝ ∂u1 ⎠ ⎝ ∂u 2 ∂u 3 ⎠
ρ ρ ρ
= h1 e1.(h 2 e 2 × h 3 e3 )
ρ ρ ρ ρ ρ
= h1 e1 . h 2 h 3 (e 2 × e3 ) =h1 e1 .h 2 h 3 e1
77
ρ ρ
= h1 h2 h3 e1.e1
ρ ρ
= h1 h2 h3 e1.e1
ρ ρ
= h1 h2 h3 [Θ e1.e1 = 1 ]
Hence proved.
ρ ρ
Contravariant Components of A and covariant components of A
ρ
A in terms of unit base vectors ê1 , ê 2 , ê 3 or Ê1 , Ê 2 , Ê 3 can be written as
ρ
A = A1 ê1 + A 2 ê 2 + A 3ê 3 = a 1Ê1 + a 2 Ê 2 + a 3 Ê 3
ρ ρ ρ ρ
∂r ∂r ∂r
Where A1, A2, A3 & a1, a2, a3 are components of A in each that , , and ∇u1, ∇u2,
∂u 1 ∂u 2 ∂u 3
ρ
∇u3 constitite reciprocal system of vectors. We can also represent A in terms of base vectors
ρ ρ ρ
∂r ∂r ∂r
, , or ∇u1, ∇u2, ∇u3 which are called unitary base vectors & are not unit vectors.
∂u 1 ∂u 2 ∂u 3
In general,
ρ ρ ρ ρ
∂r ∂r ∂r
A = C1 + C2 + C3
∂u 1 ∂u 2 ∂u 3
ρ ρ
ρ ρ ρ ρ ∂r
⇒ A = C1α1 + C 2 α 2 + C3 α 3 where α p = , p = 1,2,3
∂u p
ρ
& A = C1∇u1 + C2∇u2 + C3∇u3
ρ ρ ρ ρ
= C1 β1 + C 2β 2 + C3β 3 where β p = ∇up, p = 1, 2, 3
ρ
where C1, C2, C3 are called contravariant component of A and C1, C2, C3 are called covariant
ρ
components of A .
ρ
Example: - Let A be a given vector defined w.r.t. two gereral curvilinear coordinates system
(u1, u2, u3)& ( u1 , u 2 , u 3 ) . Find the relation between the contravariant components of the vector in
the two coordinate system.
(Find relation between Cp and Cp )
Solution: - Suppose the transformation equation from rectangular (x, y, z) system to the two
general curvilinear coordinations systems (u1, u2, u3) and (u1, u2 , u3 ) are given by
x = x1(u1, u2, u3), y = y1(u1, u2, u3), z = z1(u1, u2, u3)
78
x = x2 (u1, u 2 , u3 ) , y = y2 (u1, u2 , u3 ) , z = z2 (u1, u 2 , u3 ) …(1)
Then
I a transformation directly from (u1,u2, u3) system to (u1, u 2 , u3 ) system is defined by
∂u 1 ∂u ∂u
⇒ du1 = du 1 + 1 du 2 + 1 du 3
∂u 1 du 2 du 3
∂u 2 ∂u ∂u
du2 = du1 + 2 du 2 + 21 du 3
∂u 1 du 2 du 3
∂u 3 ∂u ∂u
du3 = du 1 + 3 du 2 + 3 du 3
∂u1 du 2 du 3
ρ ρ ρ ⎛ ∂u ∂u ∂u ⎞
α1du1 + α 2 du 2 + α 3 du 3 = ⎜⎜ 1 du1 + 1 du 2 + 1 du 3 ⎟⎟ α1
⎝ ∂u1 du 2 du 3 ⎠
79
⎛ ∂u ∂u ∂u ⎞ρ
+ ⎜⎜ 2 du1 + 2 du 2 + 21 du 3 ⎟⎟α 2
⎝ ∂u1 du 2 du 3 ⎠
⎛ ∂u ∂u ∂u ⎞ρ
+ ⎜⎜ 3 du1 + 3 du 2 + 3 du 3 ⎟⎟α 3
⎝ ∂u1 du 2 du 3 ⎠
ρ ρ ρ
= α 1d u 1 + α 2 d u 2 + α 3 d u 3
ρ ρ ∂u ρ ∂u ρ ∂u
We get α1 = α1 1 + α 2 2 + α 3 3
∂u1 ∂u1 ∂u1
ρ ρ ∂u ρ ∂u ρ ∂u
α 2 = α1 1 + α 2 2 + α 3 3
∂u 2 ∂u 2 ∂u 2
ρ ρ ∂u ρ ∂u ρ ∂u
α 3 = α1 1 + α 2 2 + α 3 3 …(6)
∂u 3 ∂u 3 ∂u 3
ρ ρ ρ ρ
Also A = α1C1 + α 2 C 2 + α 3 C3
ρ ρ ρ
= α1 C1 + α 2 C2 + α 3 C3 …(7)
ρ
where C1, C2, C3 and C1 , C2 , C3 are contravariant components of A in the two systems (u1, u2,
ρ ρ ρ ρ ⎛ ρ ∂u ρ ∂u ρ ∂u ⎞
C1 α1 + C 2 α 2 + C3α 3 = C1 ⎜⎜ α1 1 + α 2 2 + α 3 3 ⎟⎟
⎝ ∂u1 ∂u1 ∂u1 ⎠
ρ ⎛ ρ ∂u ρ ∂u ρ ∂u ⎞
+ C 2 ⎜⎜ α1 1 + α 2 2 + α 3 3 ⎟⎟
⎝ ∂u 2 ∂u 2 ∂u 2 ⎠
ρ ⎛ ρ ∂u ρ ∂u ρ ∂u ⎞
+ C3 ⎜⎜ α1 1 + α 2 2 + α 3 3 ⎟⎟
⎝ ∂u 3 ∂u 3 ∂u 3 ⎠
ρ ⎛ ρ ∂u ∂u ∂u ⎞
= α1 ⎜⎜ C1 1 + C2 1 + C3 1 ⎟⎟
⎝ ∂u1 ∂u 2 ∂u 3 ⎠
ρ ⎛ ρ ∂u ∂u ∂u ⎞
+ α 2 ⎜⎜ C1 2 + C2 2 + C3 2 ⎟⎟
⎝ ∂u1 ∂u 2 ∂u 3 ⎠
ρ ⎛ ρ ∂u ∂u ∂u ⎞
+ α 3 ⎜⎜ C1 3 + C2 3 + C3 3 ⎟⎟
⎝ ∂u1 ∂u 2 ∂u 3 ⎠
ρ ρ ρ
Equating coefficients of α1 , α 2 , α 3 , we get
80
∂u 1 ∂u ∂u
C1 = C1 + C 2 1 + C3 1
∂u 1 ∂u 2 ∂u 3
∂u 2 ∂u ∂u
C2 = C1 + C 2 2 + C3 2
∂u 1 ∂u 2 ∂u 3
∂u 3 ∂u ∂u
C3 = C1 + C2 3 + C3 3 …(8)
∂u1 ∂u 2 ∂u 3
∂u p ∂u p ∂u p
or Cp = C1 + C2 + C3 , p = 1, 2, 3 …(9)
∂u1 ∂u 2 ∂u 3
3 ∂u p
or Cp = ∑ Cq ∂u , p = 1, 2, 3 …(10)
q =1 q
Equation (8), (9), (10), (11) gives the relation between contravariant components of vector in two
coordinates systems.
ρ
Example: - Let A be given vector defined w.r.t. two curvilinear coordinates system (u1, u2, u3)
and (u1, u 2 , u3 ) . Find the relation between the covariant components of the vectors in the two co
ord system.
ρ
Solution: - Let the covariant component of A in the system (u1, u2, u3) and (u1, u 2 , u3 ) are C1, C2,
C3 and C1 , C2 , C3 respectively
ρ
∴ A = C1∇u1 + C2∇u2 + C3∇u3 = C1∇u1 + C2 ∇u 2 + C3∇u 3 …(1)
Since u p = u p ( u 1 , u 2 , u 3 ) , p=1, 2, 3
∂u p ∂u p ∂u1 ∂u p ∂u 2 ∂u p ∂u 3
= + +
∂x ∂u 1 ∂x ∂u 2 ∂x ∂u 3 ∂x
∂u p ∂u p ∂u1 ∂u p ∂u 2 ∂u p ∂u 3
= + +
∂y ∂u 1 ∂y ∂u 2 ∂y ∂u 3 ∂y
∂u p ∂u p ∂u1 ∂u p ∂u 2 ∂u p ∂u 3
= + + …(2)
∂z ∂u 1 ∂z ∂u 2 ∂z ∂u 3 ∂z
81
⎛ ∂u ∂u ∂u ⎞
Also C1∇u1 + C2∇u2 + C3∇u3 = C1 ⎜⎜ 1 î + 1 ĵ + 1 k̂ ⎟⎟
⎝ ∂x ∂y ∂z ⎠
⎛ ∂u ∂u ∂u ⎞
+ C2 ⎜⎜ 2 î + 2 ĵ + 2 k̂ ⎟⎟
⎝ ∂x ∂y ∂z ⎠
⎛ ∂u ∂u ∂u ⎞
+ C3 ⎜⎜ 3 î + 3 ĵ + 3 k̂ ⎟⎟
⎝ ∂x ∂y ∂z ⎠
⎛ ∂u ∂u ∂u ⎞
= ⎜ C1 1 + C 2 2 + C 3 3 ⎟ î
⎝ ∂x ∂x ∂x ⎠
⎛ ∂u ∂u ∂u ⎞
+ ⎜⎜ C1 1 + C 2 2 + C 3 3 ⎟⎟ ĵ
⎝ ∂y ∂y ∂y ⎠
⎛ ∂u ∂u ∂u ⎞
+ ⎜ C1 1 + C 2 2 + C 3 3 ⎟ k̂ …(3)
⎝ ∂z ∂z ∂z ⎠
⎛ ∂u ∂u ∂u ⎞
and C1∇u1 + C 2 ∇u 2 + C3∇u 3 = ⎜ C1 1 + C2 2 + C3 3 ⎟ î
⎝ ∂x ∂x ∂x ⎠
⎛ ∂u ∂u ∂u ⎞
+ ⎜⎜ C1 1 + C2 2 + C3 3 ⎟⎟ ĵ
⎝ ∂y ∂y ∂y ⎠
⎛ ∂u ∂u ∂u ⎞
+ ⎜ C1 1 + C2 2 + C3 3 ⎟ k̂ …(4)
⎝ ∂z ∂z ∂z ⎠
∂u 1 ∂u ∂u ∂u ∂u ∂u
C1 + C 2 2 + C 3 3 = C1 1 + C2 2 + C3 3 …(5)
∂z ∂z ∂z ∂z ∂z ∂z
Substituting equation (2) with p = 1, 2, 3 on R.H.S in the corresponding equation of (5) and
∂u 1 ∂u 2 ∂u 3 ∂u ∂u ∂u ∂u ∂u ∂u
equating coefficients of , , or 1 , 2 , 3 or 1 , 2 , 3 on each side,
∂x ∂x ∂x ∂y ∂y ∂y ∂y ∂y ∂y
We can get,
82
Taking first equation of (5),
∂u 1 ∂u ∂u ⎡ ∂u ∂u ∂u ∂u ∂u ∂u ⎤
C1 + C 2 2 + C 3 3 = C1 ⎢ 1 1 + 1 2 + 1 3 ⎥
∂x ∂x ∂x ⎣ ∂u 1 ∂x ∂u 2 ∂x ∂u 3 ∂x ⎦
⎡ ∂u ∂u ∂u ∂u ∂u ∂u ⎤
+ C1 ⎢ 2 1 + 2 2 + 2 3 ⎥
⎣ ∂u 1 ∂x ∂u 2 ∂x ∂u 3 ∂x ⎦
⎡ ∂u ∂u ∂u ∂u ∂u ∂u ⎤
+ C1 ⎢ 3 1 + 3 2 + 3 3 ⎥
⎣ ∂u1 ∂x ∂u 2 ∂x ∂u 3 ∂x ⎦
∂u1 ∂u 2 ∂u 3
Here equating coefficients of , ,
∂x ∂x ∂x
We get
∂u1 ∂u ∂u
C1 = C1 + C2 2 + C3 3
∂u1 ∂u 1 ∂u1
∂u1 ∂u ∂u
C2 = C1 + C2 2 + C3 3
∂u 2 ∂u 2 ∂u 2
∂u1 ∂u ∂u
C3 = C1 + C2 2 + C3 3 …(6)
∂u 3 ∂u 3 ∂u 3
∂u1 ∂u ∂u
or Cp = C1 + C2 2 + C3 3 , p = 1, 2, 3
∂u p ∂u p ∂u p
3 ∂u q
or Cq = ∑ Cq ∂u p
, p = 1, 2, 3
q =1
ρ ρ ρ
ρ ρ ρ ⎛ ∂r ∂r ∂r ⎞
Solution: - (ds) m = d r = d r .d r = ⎜⎜
2
du1 + du 2 + ⎟
⎝ ∂u1 ∂u 2 ∂u 3 ⎟⎠
83
ρ ρ ρ
⎛ ∂r ∂r ∂r ⎞
. ⎜⎜ du1 + du 2 + ⎟
⎝ ∂u1 ∂u 2 ∂u 3 ⎟⎠
ρ ρ ρ ρ ρ ρ
(ds)2 = (α1du1 + α 2 du 2 + α 3 du 3 ).(α1du1 + α 2 du 2 + α 3 du 3 )
ρ
ρ ∂r
where α p =
∂u p
ρ ρ ρ ρ ρ ρ
= α1.α1du12 + α1.α 2 du1du 2 + α1.α 3du1du 3
ρ ρ ρ ρ ρ ρ
+ α 2 .α1du 2 du1 + α 2 .α 2 du 22 + α 2 .α 3du 2 du 3
ρ ρ ρ ρ ρ ρ
+ α 3 .α1du 3du1 + α 3 .α 2 du 3du 2 + α 3 .α 3du 2 du 32
3 3
ρ ρ
(ds)2 = ∑ ∑ g pq du p du q , g pq = α p .α q
q =1 q =1
Thos is called fundamental quadratic form or Metric form. The quantities gpq are called
metric coefficients and these are symmetric i.e. = gpq = gqp
If gpq = 0 p ≠ q , the coordinate system is orthogonal.
∂x ∂x ∂y ∂y ∂z ∂z
= + + , p = 1, 2, 3
∂u p ∂u q ∂u p ∂u q ∂u p ∂u q
84
∂x ∂y ∂z ∂x ∂y ∂z
∂u 1 ∂u1 ∂u 1 ∂u1 ∂u1 ∂u1
ρ ρ ρ 2
⎛ ∂r ∂r ∂r ⎞ ∂x ∂y ∂z ∂x ∂y ∂z
⎜⎜ . . ⎟⎟ =
⎝ ∂u 1 ∂u 2 ∂u 3 ⎠ ∂u 2 ∂u 2 ∂u 2 ∂u 2 ∂u 2 ∂u 2
∂x ∂y ∂z ∂x ∂y ∂z
∂u 3 ∂u 3 ∂u 3 ∂u 3 ∂u 3 ∂u 3
∂x ∂y ∂z ∂x ∂x ∂x
∂u 1 ∂u1 ∂u1 ∂u 1 ∂u 2 ∂u 3
∂x ∂y ∂z ∂y ∂y ∂y
=
∂u 2 ∂u 2 ∂u 2 ∂u 1 ∂u 2 ∂u 3
∂x ∂y ∂z ∂z ∂z ∂z
∂u 3 ∂u 3 ∂u 3 ∂u 1 ∂u 2 ∂u 3
Example (b) show that the volume element in general coordinate is g du1 du2 du3.
Solution: - The volume element is given by
ρ ρ ρ
⎛ ∂r ⎞ ⎛ ∂r ⎞ ⎛ ∂r ⎞
dV = ⎜⎜ du 1 ⎟⎟ .⎜⎜ du 2 ⎟⎟ × ⎜⎜ du 3 ⎟⎟
⎝ ∂u 1 ⎠ ⎝ ∂u 2 ⎠ ⎝ ∂u 3 ⎠
ρ ρ ρ
∂r ∂r ∂r
= . × du1 du2 du3
∂u 1 ∂u 2 ∂u 3
85
LESSON 4 RANDOM VARIABLE AND MATHEMATICAL
EXPECTATION
Sample space
The set points representing the possible outcomes of an experiment is called the sample space of
the experiment.
Example: - (1) In tossing one coin, the sample space is S = {H, T}
(2) Two coins are tossed,
S = {HH, HT, TH, TT}
(3) In throw a dice,
S = {(1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6)
(2, 1), (2, 2), (2, 3), (2, 4), (2, 5), (2, 6)
Μ
(6, 1), (6, 2), (6, 3), (6, 4), (6, 5), (6, 6)
Total outcome = 36
Rondom Variable
A random variable is a real valued function defined on a sample space. When an experiment is
performed, several outcomes are possible corresponding to each outcome, a number can be
associatd.
Example: - If two coins are tossed, the possible outcomes are TT, TH, HT, HH
Let X → denotes the number of heads
The number associated with the outcome are : TT, TH, HT, HH
No. of heads, X : 0, 1, 1, 2
The variable X is said to be random variable.
and may be defined as
“ Let S be the sample space associated with a given experiment. A real valued function defined
on S & taking values in R( −∞, ∞) ( real no.) is called a random variable (or chance variable or
Stochastic variable or variate).
86
Discrete and Continouse Sample Space
A sample space that consists of a finite number or an infinite sequence, of points is called a
discrete sample space and that consints of one or more intervals of points is called a cotinouse
sample space.
Discrete and Continouses Random Variable
A random variable defined on a discrete sample space is called discrete r.v. or If a r.v. takes at
most a countable no. values, it is called Discrete r.v.
A r.v. is said to be Continuous if it can takes al possible value between certain limits or certain
interval.
Example: -
1) If X represents the sum of points on two dice, X is a discrete r.v.
2) If X represents the height or weight of students in a class , then it is a continuous
r.v.
3) If X represents the amount of rainfall, it is a continuous r.v.
Density function (d.f) or probability density function(p.d.f)
A function associated with discrete r.v.
X s.t. f(x) = prob[X = x] is called density function of X.
Example: - In tossing twocoins,
Outcomes = {HH, TH, HT, TT}
X = [0, 1, 2 ]
1
f(0) = P[X = 0] =
4
f(1) = P[X = 1] = 2/4 = ½
f(2) = P[X = 2] = ¼
Example: - In throwing two dice, the sample space of sum of points on two dice is
S = [1, 2, 3,……,12]
1
f(2) = P[X = 2] = (1, 1) = = f(12) (6,6)
36
2
f(3) = = f(11) [Θ (2, 1), (1, 2)]
36
3
f(4) = f(10) =
36
87
4
f(5) = f(9) =
36
5
f(6) = f(8) =
36
6
f(7) = P[X = 7] =
36
Also ∑ f (x i ) = 1
X = xi = 2, 3,……12
Distribution function
For a r.v. X, the function F(x) = P(X ≤ x) is called the distribution function of X or Cumulative
distribution.
Since F(x) = P(X ≤ x), then f(x) the p.d.f.
We have
F(x) = ∑ f (t)
t ≤x
88
71
(3) P(X ≥ 3) = P(3) + P(4) + … + P(8) =
80
9 71
Also P(X ≥ 3) = 1− P(X < 3) = 1− =
80 80
(4) P(0 < X < 5) = P(1) + P(2) + P(3) + P(4) = 25/80
Distribution function F(x) is obtained
1 4 9 17 26 37 49 63 80
F(x) : =1
80 80 80 80 80 80 80 80 80
Joint density function
Let X, Y be two r.v. , Joint d.f. gives the probability that X will assume Y will taken a value y
i.e. f(x, y) = P(X = x, Y = y)
Example: - 52 cards and 2 cards are drawn
X → no. of spade in the 1st draw
Y → no. of spade in the 2nd draw
Without replacing the 1st card drawn.
39 38
f(0, 0) = P[X = 0, Y = 0] = ×
52 51
39 13
f(0, 1) = P[X = 0, Y = 1] = ×
52 51
Y 0 1 Total
X
0 19/34 13/68 51/68
1 13/68 1/17 17/68
Total 51/68 17/68 1
13 39
f(1, 0) = P[X = 1, Y = 0] = ×
52 51
13 12
f(1, 1) = P[X = 1, Y = 1] = ×
52 51
Let A & B be two events, then
P(A ∩ B)
P(B/A) =
P( A)
89
P(A ∩ B)
P(A/B) =
P(B)
Conditional density function
f(y/x) is defined as
f(y/x) = P[Y = y | X = x]
gives distribution of Y when X is fixed
f ( x , y)
f(y/x) =
f (x)
f ( x , y)
and f(x/y) =
f (x)
Marginal density function
Here f(x, y) = f(y/x).f(x) …(1)
Condidional distance Y = y when X = x is fixed.
∑ f (y / x) = 1
y
∑ f ( x , y) = ∑ f ( y / x ) f ( x )
y y
⇒ ∑ f ( x , y) = f ( x )
y
90
2 4
f(1,0) = f(1) f(0/1) = .
6 5
2 1
f(1, 1) = f(1) f(1/1) = .
6 5
Marginal density function of X is
f(x) = ∑ f ( x, Y)
y
6 4 10 2
⇒ f0) = + = =
15 15 15 3
& f(1) = ∑ f (1, y)
y
8 2 10
f(1, 0) + f(1, 1) = + =
30 30 30
1
f(1) = …(2)
3
The conditional density function of Y for fixed x can be obtained from (1),
f ( x , y)
f(y/x) =
f (x)
f (1,0) 4 / 15 4
∴ f(0/1) = = = [using (1)]
f (1) 1/ 3 5
f (1,1) 1 / 15 1
f(1/1) = = =
f (1) 1/ 3 5
Continouse r.v. :- −∞ < X < ∞
Density function
A d.f. for a continuouse r.v. X is a function f(x) that possesses the following properties
(i) f(x) ≥ 0
∞
(ii) ∫ f ( x ) dx =1
−∞
b
(iii) ∫ f ( x ) dx = P[a < X < b] where a < b
a
91
Distribution function
∞
F(x) = ∫ f (x) dx = P[X ≤ x]
−∞
∞
f(y) = ∫ f (x, y) dx → Marginal density function of Y.
−∞
1) Determine k
2) Examine whether X&Y are independent.
Solution: - 1) For d.f. ∫∫ f ( x, y) dx dy = 1
11
∫ ∫ (xy + e
x
∴ k ) dx dy = 1
00
1 1
⎛ x2 ⎞
⇒ k ∫ ⎜⎜ y + e x ⎟⎟ dy = 1
0⎝
2 ⎠0
1
⎛y ⎞
⇒ k ∫ ⎜ + e − 1⎟ dy = 1
0⎝
2 ⎠
1
⎡ y2 ⎤
⇒ k ⎢ + ey − y ⎥ = 1
⎣4 ⎦0
92
⎡1 ⎤
⇒ k ⎢ + e − 1⎥ = 1
⎣4 ⎦
⎛ 3⎞
⇒ k⎜ e − ⎟ = 1
⎝ 4⎠
4
⇒ k=
4e − 3
1
(2) f(x) = ∫ f ( x , y) dy
0
1
f(y) = ∫ f ( x , y) dx
0
1
⎡ xy 2 ⎤
= k⎢ + e x y⎥
⎣ 2 ⎦0
⎡x ⎤
= k⎢ + ex ⎥
⎣2 ⎦
1
f(y) = k ∫ ( xy + e x ) dx
0
1
⎡x2 ⎤
= k⎢ y + ex ⎥
⎣ 2 ⎦0
⎡y ⎤
= k ⎢ + e − 1⎥
⎣2 ⎦
So f(x,y) ≠ f(x) f(y) so they are not independent.
⎧2 , 0 < x < 1, 0 < y < x
Example: - f(x, y) = ⎨
⎩0 , otherwise
1) Find marginal d.f. of X and Y
2) Find conditional d.f. of Y given X = x
Find conditional d.f. of X given Y = y
93
3) Check whether X&Y are independent or not?
x
Solution: -(1) f(x) = ∫ f ( x , y) dy
0
x
= ∫ 2 dy = (2 y )0x = 2x, 0 < x < 1
0
1 1
f(y) = ∫ f ( x, y) dx = 2 ∫ dx = (2 y )y
1
y y
a
1 1
P(X ≤ a) = ⇒ ∫ f ( x ) dx =
2 0
2
a
a
1 ⎛ x3 ⎞ 1
⇒ 3 ∫ x dx = ⇒ a = ⎜⎜ ⎟⎟ =
2
0
2 ⎝ 3 ⎠0 2
1/ 3
3 1 ⎛1⎞
⇒ a = ⇒a =⎜ ⎟
2 ⎝2⎠
1 1
1
∫ f (x ) dx = ∫ 3x
2
Also dx =
a a
2
94
1
⎛ x3 ⎞ 1 1
⇒ 3 ⎜⎜ ⎟⎟ = ⇒ 1 − a 3 =
⎝ 3 ⎠a 2 2
1/ 3
1 ⎛1⎞
⇒ a3 = ⇒a =⎜ ⎟
2 ⎝2⎠
1
(2) P(X > b) = 0.05 ⇒ ∫ f ( x ) dx = 0.05
b
1
1
⎛ x3 ⎞
⇒ 3 ∫ x dx = 0.5
2
⇒ 3 ⎜⎜ ⎟⎟ = 0.05
b ⎝ 3 ⎠b
1 19
⇒ (1−b3) = ⇒ b3 =
20 20
1/ 3
⎛ 19 ⎞
⇒ b= ⎜ ⎟
⎝ 20 ⎠
1 2 3
⇒ 0 + ∫ f ( x ) dx + ∫ f ( x ) dx + ∫ f ( x ) dx + 0 = 1
0 1 2
1 2 3
⇒ ∫ ax dx + ∫ a dx + ∫ (3a − ax) dx =1
0 1 2
1 3
⎛ x2 ⎞ ⎛ x2 ⎞
⇒ a ⎜⎜ ⎟⎟ + a ( x )12 + a ⎜⎜ 3x − ⎟ =1
⎟
2
⎝ ⎠0 ⎝ 2 ⎠2
95
a ⎛ 9 ⎞
⇒ + a + a + ⎜9 − − 6 + 2⎟ = 1
2 ⎝ 2 ⎠
3a a
⇒ + =1
2 2
1
⇒ a=
2
1. 5
(2) Now P(X ≤ 1.5) = ∫ f (x) dx
−∞
0 1 1.5
= ∫ f dx + ∫ f ( x ) dx + ∫ f ( x ) dx
−∞ 0 1
1 3/ 2
= 0 + ∫ ax dx + ∫ a dx
0 1
⎛1⎞ a ⎛3 ⎞
= a ⎜ ⎟ + a (x ) 13 / 2 = + a ⎜ − 1⎟
⎝2⎠ 2 ⎝2 ⎠
Example: - From the given bivariate probability distribution
(1) Obtain marginal distribution of X & Y.
(2) The conditional distribution of X given Y = 2
X −1 0 1 ∑ f ( x , y) = f ( y)
x
Y
96
f(0) = P(X = 0) = 5/15
f(1) = P(X = 1) = 4/15
Marginal distribution of Y
From above table f(y) = ∑ f ( x , y)
x
97
X→ represents the number of heads is a r. v., then total outcomes are
{HHH, HTH, HHT, THH, THT, TTH, HTT, TTT}
Here X can take values X = 0, 1, 2, 3
1 3 3 1
with f(x) = , , ,
8 8 8 8
1 3 3 1
Then E(X) = ∑ xf ( x ) = 0. + 1. + 2. + 3.
8 8 8 8
12 3
= = × 1 Rs. = Rs. 1.50
8 2
If g(x) = x2
Then g(x) = 02, 12, 22, 32
∴ E [g(x)] = ∑ g( x i )f ( x i )
1 3 3 1
= 0 2 × + 12 × + 2 2 × + 32 ×
8 8 8 8
3 12 9
= + + = Rs. 3.0
8 8 8
Mathematical Expectation for Continuous r.v.
Let X be a continuous r.v. with p.d.f. f(x), then its mathematical expectation is
∞
E(x) = ∫ x f ( x ) dx
−∞
∞
For function g(x), E[g(x)] = ∫ g ( x )f ( x ) dx
−∞
∞
=C ∫ xf (x ) dx = C E(X)
−∞
∞
⇒ E(a) = a ∫ f ( x ) dx = a.1 = a
−∞
= ∑∑ x i f (x i , y j ) + ∑∑ y jf (x i , y j )
i j i j
= ∑ x i ∑ (x i , y j ) + ∑ y j ∑ f (x i , y j )
i j j i
= ∑ x i f ( x i ) + ∑ y jf ( y j )
i j
99
∞ ∞ ∞ ∞
= ∫ ∫ xf (x, y) dx dy + ∫ ∫ yf (x, y) dx dy
−∞ −∞ −∞ −∞
∞ ⎡∞ ⎤ ∞ ⎡∞ ⎤
= ∫ ⎢∫
x ⎢ f ( x , y ) dy ⎥ dx + ∫ ⎢∫
y ⎢ f ( x , y ) dx ⎥ dy
−∞ ⎣ −∞ ⎦⎥ −∞ ⎣ −∞ ⎦⎥
∞ ∞
= ∫ xf (x) dx + ∫ yf ( y) dy
−∞ −∞
∞ ∞
= ∫ axf ( x ) dx + ∫ bf ( x ) dx
−∞ −∞
∞ ∞
= a ∫ xf ( x ) dx + b ∫ f ( x ) dx
−∞ −∞
⎡ ∞ ⎤
⇒ E(aX + b) = aE(X) + b ⎢Θ ∫ f ( x )dx = 1⎥
⎢⎣ −∞ ⎥⎦
100
Example: - If f(x, y) = e−(x + y) is the joint p.d.f. of r.v. X and Y, then find
P = (1 < X < 2, 0 < Y < 2)
∫ ∫ f (x, y) dx dy = 1
− ∞ −∞
∞ ∞ ∞ ∞
∫ ∫ e −( x + y ) dx dy = ∫ ∫e
−x −y
L.H.S = e dx dy
−∞ −∞ −∞ −∞
∞
= ∫ [− e −x −y ∞
e −∞ ] dy
−∞
∞
= ∫ P(1 < X < 2,0 < Y < 2)
−∞
2 2
∫ ∫e
−( x + y )
= dy dx
1 0
2 2
∫ ∫e
−x − y
= e dy dx
1 0
2 2 2
= ∫ e − x dx ∫ e − y dy = (1 − e −2 ) ∫ e − x dx
1 0 1
⎛ 1 ⎞ ⎛1 1 ⎞
= (1 − e−2) (e−1 − e−2) = ⎜1 − 2 ⎟ ⎜ − 2 ⎟
⎝ e ⎠ ⎝e e ⎠
⎛ e2 −1⎞ ⎛ e −1⎞
= ⎜⎜ 2 ⎟⎟ ⎜ 2 ⎟
⎝ e ⎠⎝ e ⎠
(e 2 − 1)(e − 1) (e 3 − e 2 − e + 1)
= =
e4 e4
⎧C( x 1 x 2 + e x1 ), 0 < ( x 1 , x 2 ) < 1
Example: - Let f(x1, x2) = ⎨
⎩0 , otherwise
(i) Determine C
(ii) Examine whether X1 and X2 are independent or not.
101
Solution: - For density function F(x1, x2) we must have
∞ ∞
∫ ∫ f (x, y) dx dy = 1
−∞ −∞
1
⎡ x 12 x 2
1
⎤
⇒ C ∫⎢ + e x1 ⎥ dx2 = 1
0⎣
2 ⎦0
1
⎡1 ⎤
⇒ C ∫ ⎢ x 2 + e − 1⎥ dx2 = 1
0⎣
2 ⎦
1
⎡1 x2 ⎤
⇒ C ⎢ × 2 + ex 2 − x 2 ⎥ = 1
⎣2 2 ⎦0
⎡1 ⎤
⇒ C ⎢ + e − 1⎥ = 1
⎣ 4 ⎦
⇒ C [1 + 4e − 4] = 4
4
or C=
4e − 3
Theorem: - Show that the expectation of the product of independent r.v. is equal to product of
their expectations. i.e. E(XY) = E(X).E(Y)
Proof: - Let X and Y are two independent random variables, then
X: x1, x2, …….,xn
With d.f: - f(x1), f(x2), …..f(xn)
and Y : y1, y2, ……, ym
with d.f: f(y1), f(y2), ……f(ym)
n m
∴ E(X) = ∑ x i f ( x i ), E ( Y ) = ∑ y jf ( y j )
i =1 j =1
102
n m
= ∑ ∑ x i y j f ( x i )f ( y j )
i =1 j=1
= ∑ x i f ( x i )∑ y jf ( y j )
i j
∞ ∞
E(XY) = ∫ ∫ xyf (x )f ( y) dx dy
−∞ −∞
∞ ∞
= ∫ xf (x ) dx ∫ yf ( y) dy = E(X) E(Y)
−∞ −∞
In general, we get
E(X1 X2 X3…..Xn) = E(X1) E(X2)……E(Xn)
Example: - Let X represents the number on the face of dice then
X: 1 2 3 4 5 6
1 1 1 1 1 1
f(x):
6 6 6 6 6 6
Now E(X) = ∑ xf (x)
1 21 7
= (1 + 2 + ……. + 6) = =
6 6 2
And when X is sum of points when two dies are thrown, i.e.
X: 2 3 4 5 6 7 8 9 10 11 12
f(x): 1 2 3 4 5 6 5 4 2 2 1
36 36 36 36 36 36 36 36 36 36 36
1
∴ E(X) = ∑ xf ( x ) = 36 [2 + 6 + 12 + 20 + 30 + 42 + 40 + 36 + 30 + 22 + 12]
1
⇒ E(X) = . 252 = 7
36
103
Example: - Let X be the profit that a person makes in a business. He may earn Rs. 2800 with a
probability 0.5, he may lose Rs 5500 with probability 0.3 and he may neither earn nor lose with a
probability 0.2. Calculate E(X)
Solution: - Here P(X = 2800) = 0.5
P(X = −5500) = 0.3
P(X = 0) = 0.2
Then E(X) = ∑ xf (x)
= 2800(0.5) + (−5500) (0.3) + (0) (0.2)
= 1400 − 1650 = − 250, he may lose Rs 250.
Example: - A and B in turns throw an ordinary dice for a price of Rs. 44. The first to throw a
“six” wins. If A has first throw, what is his expectation?. Also calculate B’s expectation
1
Solution: - The problem of getting a “six” on dice is p(x) =
6
A has 1st throw, so he can win in the 1st, 3rd, 5th ……
Hence A′ chance (probability) of winning is
1 5 5 1 5 5 5 5 1
= + × × + × × × × ……
6 6 6 6 6 6 6 6 6
1 ⎡ ⎛5⎞ ⎛5⎞ ⎤
2 4
= ⎢1 + ⎜ ⎟ + ⎜ ⎟ + ......⎥
6 ⎣⎢ ⎝ 6 ⎠ ⎝ 6 ⎠ ⎦⎥
1 1 1 36 6
= . = . =
2
6 ⎛5⎞ 6 11 11
1− ⎜ ⎟
⎝6⎠
6
∴ Amount of A = 44× = Rs. 24
11
Similarly B can win in 2nd, 4th, 6th,…..
Hence B′(chance) of winning are
5 1 5 5 5 1
= . + . . . + ……
6 6 6 6 6 6
5 1 ⎡ ⎛5⎞ ⎤
2
= . ⎢1 + ⎜ ⎟ + ......⎥
6 6 ⎢⎣ ⎝ 6 ⎠ ⎥⎦
104
1 5 36 5
= . . =
6 6 11 11
5
∴ Amount of B = × 49 = Rs. 20
11
Example: - A bag contains a coin of value M and a number of other coins whose aggregate
value is m. a person draws one at a time till he draws the coin M, find the value of his
expectation.
Solution: - Let there be K other coins each of value m/K, so that their aggregate value is m. he
may draw the coin M at 1st draw or 2nd or 3rd or…..or (K+1)th draw with probability
1 1 1
, , , ……
K +1 K +1 K +1
1 ⎡ 1 ⎤ 1 ⎡ 1 ⎤⎡ 1⎤ 1
Θ , ⎢1 − , ⎢1 − 1− ,.......
K + 1 ⎣ K − 1⎦ K ⎣ K + 1⎥⎦ ⎢⎣ K ⎥⎦ K − 1
⎥
1 1 1
= , , , ……….
K +1 K +1 K +1
The corresponding amount drawn X is
m 2m (k − 1)m Km
M, + M, + M,......., + M, +M
K K K k
1 ⎡ m 2m Km ⎤
∴ E(X) = ⎢ M+ +M+ + M + ....... + + M⎥
K +1 ⎣ K K K ⎦
1 ⎡ m ⎤
= ⎢ M + (K + 1) + (1 + 2 + 3 + ...... + K )⎥
K +1 ⎣ K ⎦
1 m K (K + 1) m
=M+ . =M+
K +1 K 2 2
105
LESSON 5 MOMENTS AND MOMENT GENERATING FUNCTIONS
Moments: - Let X is a r. v., then E [Xr], if exists is called the rth moment of X about origin and is
denoted by μ ′r
Where μ = E(X)
and f(xi) = P(X = xi)
when r = 1 from (1), we get
μ ′r = E(X) = ∑ xf (x) = Mean
and from (2),
μ1 = E (X − μ) = E (X) − E(μ)
⇒ μ1 = μ − μ = 0 [Θ E(μ) = μ, E(X) = μ]
Moments about Mean (μr) in terms of moments about any point ‘a’:-
Let X → r. v.,
E(X) = X = μ
then μr = E [X − μ]r = E [X − a − μ + a]r
⇒ μr = E(X − a − d)r
Where d = μ − a
or μr = E [(X − a)r − rC1d(X − a)r − 1 + …..+ rCr − 1(−1)r − 1dr − 1(X − a) + dr(−1)r]
⇒ μr = μ ′r (a) − rC1 d μ ′r −1 (a) + ……+rCr − 1(−1)r − 1dr − 1E(X − a) + dr(−1)r
106
Now μ1 = 0 [μ1 = μ1′ (a) − d = μ1′ (a) − μ1′ (a) = 0]
If X is a continuous r.v.
∞
μ 'r = ∫ ( x − a ) r f ( x ) dx
−∞
∞
μr = ∫ ( x − μ) r f ( x ) dx
−∞
Also μ2 = E(X − μ)2. This E(X − μ)2 is called Variance and is denoted by σ2.
Covariance between X and Y
Cov(X, Y) = E [(X− X ) (Y − Y )]
= E [(X −E(X)) (Y −E(y))]
Example: - Find E(X), E(X2), E(X − E(X))2 from the following:-
X: 8 12 16 20 2
1 1 3 1 1
f(x):
8 6 8 4 12
Solution: - E(X) = ∑ xf ( x )
1 1 1
= 8 × + 12 × + ....... + 24 ×
8 6 12
⇒ E(X) = 16 = Mean
1 1 3 1
and E(X2) = ∑ x 2 f (x ) = 64 × 8 + 144 × 6 + 256 × 8 + ...... + 576 × 12
⇒ E(X2) = 276
107
⇒ μ2 = 276 − (16)2 = 20
This μ2 = 20 is the variance.
Example: - Find E(X), E(X2), E(X − E(X))2
X: 2 3 4 5 6 7 8 9 10 11 12
f(x): 1 2 3 4 5 6 5 4 2 2 1
36 36 36 36 36 36 36 36 36 36 36
1
Solution: - E(X) = ∑ xf (x ) = 36 [2+6+12+20 + 30 + 42 + 40 + 36 + 30 + 22 + 12]
= 7 = Mean
1
E(X2) = ∑ Ex 2 f (X) = 36 [4+18+48+10+305+49+64+81+100+121+144]
1 329
= (1974) =
36 6
μ2 = E [X − E(X)]2 = E(X2) − [E(X)]2
1974 329
= − 49 = − 49
36 6
329 − 294 35
= =
6 6
Let m represents median, then
P(X < m) = P(X > m)
In this eg, m = 7
mode = 7
Also mean = 7
So it is a very good distribution.
Example: - For distribution,
⎧1 2
⎪16 (3 _ x , − 3 < x ≤ −1
⎪
⎪ 1 (6 − 2 x 2 ) , −1 < x ≤ 1
⎪
f(x) = ⎨16
⎪1 2
⎪ (3 − x ) , 1< x ≤ 3
⎪16
⎪⎩0 , otherwise
Check whether it represents a probability distribution or not. Also find its mean and variance.
108
Solution: - First we prove that
∞ 3
∫ f (x ) dx = 1 or ∫ f (x ) dx = 1
−∞ −3
∞ ⎛ −3 − 1 1 3 ∞ ⎞
⎜ + + + + ⎟ f dx
Now ∫ f dx =
⎜∫ ∫ ∫ ∫ ∫ ⎟
−∞ ⎝ − ∞ −3 −1 1 3 ⎠
∞ −1 1 3
1 1 1
⇒ ∫ f (dx ) = ∫ 16
(3 + x 2 )dx + ∫
16
(6 − 2 x 2 )dx + ∫
16
(3 − x 2 )dx
−∞ −3 −1 1
−1 1 3
1 ⎡ (3 + x ) 3 ⎤ 1 ⎡ 6 2x 3 ⎤ 1 ⎡ (3 − x ) 3 ⎤
= ⎢ ⎥ + ⎢ x − ⎥ + ⎢ ⎥
16 ⎣ 3 ⎦ − 3 16 ⎣ 3 ⎦ −1 16 ⎣ − 3 ⎦1
1 1 ⎡ 2 2⎤ 1
= [8 − 0] + ⎢6 − + 6 − ⎥ − [0 − 8]
48 16 ⎣ 3 3 ⎦ 48
1 1 ⎛ 4⎞ 1
= + ⎜12 − ⎟ +
6 16 ⎝ 3⎠ 6
∞
1 1 32 1 2
⇒ ∫ f (x)dx = 3 + 16 . 3 = + =1
3 3
−∞
∞
Also Mean = E(X) = ∫ xf (x ) dx
−∞
3
⇒ E(X) = ∫ xf (x )dx
−3
−1 1 3
x x x
= ∫ 16
(3 + x )2 dx + ∫
16
(6 − 2 x 2 )dx + ∫
16
(3 − x ) 2 dx
−3 −1 1
−1 −1
1 ⎡ (3 + x )3 ⎤ 1 (3 + x )3
3 ⎦ −3 −∫3 16
= ⎢x ⎥ − (1) dx + 0
16 ⎣ 3
1 ⎧⎪ ⎡ x (3 − x )3 ⎤ (3 − x )3 ⎫⎪
3 3
16 ⎪ ⎣ − 3 ⎦1 ∫1
+ ⎨⎢ ⎥ − (1) dx ⎬
−3 ⎪⎭
⎩
[Integrand in second integral is an odd function of x, so its value is zero]
109
−1 3
1 1 ⎡ (3 + x ) 4 ⎤ 1 1 ⎡ (3 − x ) 4 ⎤
= [(−1)8 + 0] − ⎢ ⎥ − [ 0 − 1,8] + ⎢ ⎥
48 48 ⎣ 4 ⎦ −3 48 48 ⎣ − 4 ⎦1
1 1 1 1 1 1
= − − . (16 − 0) + − . (0 − 16)
6 48 4 6 48 4
1 1 1 1 1 1
⇒ E(X) = − . .16 + . .16 = − + = 0
48 4 48 4 12 12
Also Variance = V(X) = E(X − E(X))2 = E(X2) − [E(X)]2
Now
3
E(X2) = ∫x
2
f ( x )dx
−3
−1 1 3
x2 x2 x2
= ∫ 16
(3 + x ) 2 dx + ∫
16
(6 − 2 x 2 )dx + ∫
16
(3 − x ) 2 dx
−3 −1 1
−1 1 3
x2 1 x2
⇒ E(X2) = ∫ (9 + x 2 + 6 x )dx + ∫ (6 x 2 − 2 x 4 )dx + ∫ (9 + x 2 − 6 x ) x
−3
16 −1
16 1
16
−1 1
1 ⎡ 9x 3 x 5 6x 4 ⎤ 1 ⎡ 6x 3 2x 5 ⎤
= ⎢ + + ⎥ + ⎢ − ⎥
16 ⎣ 3 5 4 ⎦ −3 16 ⎣ 3 5 ⎦ −1
3
1 ⎡9x 3 x 5 6x 4 ⎤
+ ⎢ + − ⎥
16 ⎣ 3 5 4 ⎦1
1 ⎛ 1 3 243 3 ⎞ 1 ⎛ 2 2⎞
= ⎜ − 3 − + + 81 + − × 81⎟ + ⎜ 2 − + 2 − ⎟
16 ⎝ 5 2 5 2 ⎠ 16 ⎝ 5 5⎠
1 ⎡ 243 3 1 ⎤
+ ⎢81 + − × 81 − 3 − × 81⎥
16 ⎣ 5 2 5 ⎦
1 ⎡ 1 3 243 243 ⎤ 1 ⎛ 4⎞
= 2. ⎢ − 3 − + + 81 + − ⎥ + ⎜4 − ⎟
16 ⎣ 5 2 5 2 ⎦ 16 ⎝ 5⎠
1⎡ 240 242 ⎤ 1 16
= ⎢ 78 − + + .
8⎣ 2 5 ⎥⎦ 16 5
1⎡ 242 ⎤ 1
= ⎢ − 42 + +
8⎣ 5 ⎥⎦ 5
110
1 32 1 4 1
⇒ E(X2) = . + = + =1
8 5 5 5 5
Hence Variance = E(X2) − [E(X)]2
= 1 − 02 = 1
Example: - Show that the value of cov(X, Y) for probability distribution
⎧ 1 −y / 3
⎪ e , 0≤x≤ y≤∞
f(x, y) = ⎨ 9 is 9.
⎪⎩0 , otherwise
∞
E(Y) = ∫ yf ( y)dy
−∞
∞ ∞
1 −y / 3 ⎡1 ⎤
= ∫ 9
e dy = ⎢ e − y / 3 (−3)⎥
⎣9 ⎦x
x
1 −y / 3
⇒ g(x) = e ,0 ≤ x ≤ ∞
3
Similarly Marginal density function of Y will be
y
1 y −y / 3
∫ 9e
−y / 3
h(y) = dx = e ,0≤y≤∞
0
9
∞
∴ E(X) = ∫ xf (x )dx
0
∞ 1 1∞
= ∫ x. e − x / 3 dx = ∫ x e − x / 3 dx
0 3 30
111
1 ⎡ −x / 3 ⎤
∞
= ⎢ x.e
3 ⎣⎢
[ ∞
]
(−3) 0 + 3∫ e − x / 3 dx ⎥
0 ⎦⎥
=
1
3
[ [
0 + 3 e − x / 3 (−3) ]]
∞
0 =3 … (1)
∞ ∞
1
E(Y) = ∫ y h ( y)dy = ∫ y 2 e − y / 3 dy
0
90
1 ⎡ 2 −y / 3 ⎤
∞
= ⎢y e
9 ⎢⎣
{ (−3) }
∞
0 + 6 ∫ ye − t / 3 dy⎥
0 ⎥⎦
2 ⎡ −y / 3 ⎤
∞
= ⎢ ye
3 ⎣⎢
[ ∞
]
(−3) 0 + 3∫ e − y / 3 dy⎥
0 ⎦⎥
=
2
3
[ ∞
]
× 3 e − y / 3 (−3) 0 = 2 × 3 = 6 … (2)
∞ ∞
E(XY) = ∫ ∫ xyf (x, y) dx dy
x =0 y = x
∞ ∞
1
= ∫ ∫ xy. e − y / 3 dx dy
9
x =0 y=x
1
∞ ⎡∞ ⎤
9 x∫=0 ⎢ y∫= x
−y / 3
= x ⎢ ye dy ⎥ dx
⎣ ⎥⎦
⎪⎧ − y / 3 ⎫⎪
[ ]
∞ ∞
1 ∞
= ∫ x
9 x =0 ⎪⎩
⎨ ye ( −3) x + 3 ∫ e − y / 3 dy⎬ dx
⎪⎭
x
∞
=
1
∫ (
x 3xe − x / 3 + 3(3)e − x / 3 dx
9 x =0
)
[ ]
∞
1
9 ∫0
= 3x 2 e − x / 3 + 9e − x / 3 x dx
1 ⎡ 2 −x / 3 ⎤
∞ ∞
= ⎢ ∫ x I e II dx + 3∫ e II− x / 3 x I dx ⎥
3 ⎣⎢ 0 0 ⎦⎥
1
= [54 + 3 × 9]
3
112
54 + 27 81
= = = 27
3 3
∴ Cov(X, Y) = E(XY) − E(X)E(Y)
= 27 − 3 × 6 = 27 − 18 = 9
Hence proved
1⎡ x−b ⎤
Example: - If f(x) = ⎢1 − ⎥ , x − b < a . Find mean and variance.
a⎣ a ⎦
1⎡ x−b ⎤
Solution: - f(x) = ⎢1 − ⎥
a⎣ a ⎦
b b+a
1 ⎡ ( x − b) ⎤ 1 ⎡ ( x − b) ⎤
= ∫ x. ⎢1 + dx + ∫ x. ⎢1 − dx
b −a
a⎣ a ⎥⎦ b
a⎣ a ⎥⎦
b b+a
1 ⎛ x 2 − bx ⎞ 1 ⎛ x 2 bx ⎞
a b∫−a ⎜⎝ a ∫b ⎜⎝
= ⎜ x + ⎟ dx + ⎜ x − + ⎟ dx
a ⎟⎠ a a ⎟⎠
b b
1 ⎡ x 2 x 3 bx 2 ⎤ 1 ⎡ x 2 x 3 bx 2 ⎤
= ⎢ + − ⎥ + ⎢ − + ⎥
a ⎣ 2 3a 2a ⎦ b−a a ⎣ 2 3a 2a ⎦ b+a
1 b2 b3 b 3 ⎛ b 2 + a 2 − 2ab ⎞ (b − a ) 2 b
= + 2 − 2 − ⎜⎜ ⎟−
⎟ 2
+ 2
(b − a ) 2
a 2 3a 2a ⎝ 2a ⎠ 3a 2a
1 ⎛ b 2 + a 2 + 2ab ⎞ (b + a ) 3 b(b + a ) 2 b 2 b 3 b3
+ ⎜⎜ ⎟−
⎟ + − + −
a⎝ 2 ⎠ 3a 2 2a 2 2a 3a 2 2a 2
113
b 2 b3 b 3 b 2 a 2 2ab b 2 a 2 2ab
= + 2− 2− − + + + +
2a 3a 2a 2a 2a 2a 2a 2a 2a
b 2 b3 b3
− + 2 − 2 =b
2a 3a 2a
Variance = V(X) = E[X − E(X)]2
∞
∫ ( x − b)
2 2
= E(X − b) = f ( x ) dx
−∞
b b+a
2 1 ⎛1+ x − b ⎞ 1 ⎡ (a − b) ⎤
= ∫ (x − b) . a ⎜⎝ a ⎟⎠ dx + ∫ ( x − b) 2 ⎢1 −
a⎣ a ⎥⎦
dx
b −a b
b b+a
1 ⎡ ( x − b) 3 ⎤ 1 ⎡ ( x − b) 3 ⎤
a b∫−a ⎣ a ∫b ⎣
2 2
= ⎢ ( x − b ) + ⎥ dx + ⎢ ( x − b ) − ⎥ dx
a ⎦ a ⎦
b b+a
1 ⎡ ( x − b ) 3 ( x − b) 4 ⎤ 1 ⎡ ( x − b) 3 ( x − b) 4 ⎤
= ⎢ + ⎥ + ⎢ − ⎥
a⎣ 3 4a ⎦ b−a a ⎣ 3 4a ⎦ b
1 a3 a4 a3 a4
= . − 2+ −
a 3 4a 3a 4a 2
2a 3 a 4 2a 2 a 2 a 2
= − 2 = − =
3a 2a 3 2 6
Example: - Find mean and Variance for the following distribution
⎧ ( x − 1) 3
⎪ , 1≤ x ≤ 3
f(x) = ⎨ 4
⎪0 , otherwise
⎩
∞
Solution: - Check for ∫ f (x ) dx = 1
−∞
∞ 3
( x − 1) 3
As mean = E(X) = ∫ xf ( x ) dx = ∫ .x dx
−∞ 1
4
114
3
1 13
= ∫ ( x 4 − x − 3x 3 + 3x 2 ) dx =
41 5
2
⎛2 13 ⎞
Variance = V(X) = E[X − E(X) ] = E ⎜ X − ⎟
⎝ 6⎠
∞ 2
⎛ 13 ⎞
= ∫ ⎜⎝ x − 5 ⎟⎠ f (x ) dx
−∞
3 2
1 ⎛ 13 ⎞ ( x − 1) 3
= ∫⎜x − ⎟ dx
4 1⎝ 5⎠ 4
⎧ 3 ⎫
1 ⎪⎡⎛ 13 ⎞ ( x − 1) 4 ⎤ 1 ⎛
2 2
13 ⎞ ⎪
⎥ − ∫ ⎜ x − ⎟( x − 1) dx ⎬
4
= ⎨⎢⎜ x − ⎟
4 ⎪⎢⎣⎝ 5⎠ 4 ⎥⎦ 2 1 ⎝ 5⎠ ⎪⎭
⎩ 1
1 ⎧⎪ 1 1 ⎡ ( x − 1) 6 ⎤ ⎫⎪
3 3
4 1 ⎡⎛ 13 ⎞ ( x − 1) 5 ⎤
= ⎨4 × − ⎢⎜ x − ⎟ ⎥ + × ⎢ ⎥ ⎬
4 ⎪ 25 2 ⎣⎝ 5⎠ 5 ⎦ 2 5 ⎣ 6 ⎦1 ⎪
⎩ ⎭
4 4 2 4
= − × +
25 5 5 15
4 8 4 4 4 8
= − + =− + =
25 25 15 25 15 75
Theorem: - If X is a r.v., then V(aX + b) = a2 V(X), where a and b are constants.
Proof: - Let Y = aX + b, then
E(Y) = aE(X) + b
Then V(aX + b) = V(Y) = E[Y − E(Y)]2
= E[aX + b − aE(X) − b]2
= a2E(X2) + a2[E(X)]2 − 2a2[E(X)]2
= a2E(X2) − a2[E(X)]2
= a2[E(X2) − [E(X)]2] = a2 V(X)
Cor(i) If b = 0, then V(ax) = a2 V(X)
(ii) If a = 0, then V(b) = 0
(iii) If a = 1, then V(X + b) = V(X).
115
Theorem: - Prove that V(X ± Y) = V(X) ± V(Y) ± 2 Cov(X, Y) and V(X ± Y) = V(X) +V(Y)
provided X and Y are independent r.v.
Proof: - V(X + Y) = E[(X + Y) − E(X +Y)]2
= E[(X + Y) − E(X) − E(Y)]2
= E[{X − E(X)} + {Y − E(Y)}]2
= E[X − E(X)]2 + [Y − E(Y)]2 + 2E[(X − E(X)][Y − E(Y)]
= E[X −E(X)]2 + E[Y − E(Y)]2 + 2E[{X − E(X)}{Y − E(Y)}]
= V(X) + V(Y) + 2 cov(X, Y)
Similarly V(X − Y) = V(X) + V(Y) − 2cov(X, Y)
V(X ± Y) = V(X) +V(Y) ± 2cov(X, Y)
If X and Y are independent, then cov(X, Y) = 0
∴ V(X ± Y) = V(X) + V(Y) ± 2.0
= V(X) + V(Y).
Covariance: - If X and Y are two r.v., then cov between then is defined as
Cov(X, Y) = E[[X − E(X)[Y −E(Y)]]
= E[XY − XE(Y) − YE(X) + E(X) E(Y)]
= E(XY) − E(X)E(Y) − E(Y)E(X) + E(X)E(Y)
= E(XY) − E(X)E(Y)
We can also express it as
Cov(X, Y) = E[[X −E(X)][Y − E(Y)]]
= ∑ ∑ (x
i j
i − X )( y j − Y )f ( x i , y j ) for discrete case
Cov(X, Y) = E[XY − XY − XY + X Y ]
= E(XY) − X Y − Y X + X Y
= E(XY) − X Y
If X and Y are independent r.v., then
E(XY) = E(X) E(Y)
Cov(X, Y) = E(X) E(Y) − X Y = X Y − X Y = 0
116
⇒ Cov(X, Y) = 0 if X and Y are independent.
Remark
i) Cov(ax, by) = E[(aX − E(ax))(bY − E(bY))]
= E[(aX − aE(X)) (bY − BE(Y))]
= E[a[X − E(X)] . b[Y − E(Y)]]
= ab[E(X − E(X)) (Y − E(Y))]
= ab Cov(X, Y)
⇒ Covariance is not independent of change of scale.
ii) Cov(X + a, Y + b) = E[{(X + a) − E(X − a)} {Y + b − E(Y + b)}]
= E[{(X + a) − E(X) − a} { Y + b − E(Y) − b}
= E[{X −E(X)} { Y − E(Y)}]
= Cov(X, Y)
Thus Cov(X, Y) is independent of change of origin but is not independent of change of scale.
MEAN DERIVATION FOR CONTINUOUS CASE
∞
E[|X − a|] = ∫ | x − a | f(x)dx
−∞
∞
Variance = σ 2x = E[X − E(X)]2 = ∫ (x − X)
2
f(x)dx
−∞
Absolute moment
Let X be r. v. with p.d.f f(x), then its rth absolute moment about any point a is given by
∞
∫| x − a |
r r
E[|X − a| ] = f(x) dx
−∞
117
X−a
U=
h
⎛X−a⎞ 1
Now E[U] = U = E⎜ ⎟ = [E(X) − E(a)]
⎝ h ⎠ h
1
U= ( X − a ) ⇒ X = a + hU
h
Then μr = E[X − X ]r = E[a + hU − a − h U ]r
= E[hr(U − U )r] = hr E[U − U ]r
Where E[U − U ]r = rth moment of U about mean
⎡ tX t 2 X 2 trXr ⎤
or MX(t) = E ⎢1 + + + ...... + + ....⎥
⎣ 1! 2! r! ⎦
t t2 tr
=1+ E(X) + E(X 2 ) + ...... + R (X r ) + ....
1! 2! r!
t2 tr
MX(t) = 1 + tμ 1′ + μ 2 + ...... μ ′r + ....
′
2! r!
tr
where μ ′r = coefficient of in MX(t), is the rth moment about origin.
r!
The moments can also be obtained from the M.G.F through following relations:
118
dr
μ ′r = M X (t)
dt r
t=0
d
μ1′ = M X (t)
dt
t=0
d2
μ ′2 = M X (t)
dt 2
t=0 etc.
when X is continuous r.v., then MGF about origin is given by
∞
∫e
tx
MX(t) = f(x) dx
−∞
∫e
t ( x −μ )
MX(t) = f(x) dx
−∞
∫e
t ( x −a )
MX(t) = f(x) dx
−∞
⎡ tx t 2 x 2 trxr ⎤
As MX(t) = ∑ e tx f(x) = ∑ ⎢ 1! 2!
f ( x ) 1 + + + ...... + ⎥
r! ⎦
x x ⎣
d ∞ ⎡ 2 tx 2 3t 2 x 3 rt r −1 x r −1 ⎤
∴
dt
MX(t) = ∑ f (x )⎢x + 2!
+
3!
+ ...... +
r
+ .....⎥
x =0 ⎣ ⎦
∞ ∞
d
∴
dt
MX(t) = ∑ xf (x ) =∑ xf (x ) = μ1′ = 1st moment about origin.
x =0 x =1
119
∞
i.e. ∫ f (x ) dx = 1 or not
−∞
∞ ⎛0 1 2 ∞ ⎞
⎜ + + + ⎟ f(x) dx
∫ ⎜∫ ∫ ∫ ∫
Now f ( x ) dx =
⎟
−∞ ⎝ −∞ 0 1 2 ⎠
1 2
= 0 + ∫ x dx + ∫ (2 − x ) + 0
0 1
1 2
⎡x2 ⎤ ⎡ x2 ⎤
= ⎢ ⎥ + ⎢2 x − ⎥
⎣ 2 ⎦0 ⎣ 2 ⎦1
1 1
= −0+4−2−2+
2 2
∞
1 1
⇒ ∫ f (x )dx = 2 + 2 = 1
−∞
1 2
= ∫ x.x dx + ∫ x (2 − x )dx
0 1
1 2
⎡ x 3 ⎤ ⎡ 2x 2 x 3 ⎤
= ⎢ ⎥ +⎢ − ⎥
⎣ 3 ⎦0 ⎣ 2 3 ⎦1
1 8 1 2 8
= + 4 − −1_ = + 3 − = 1
3 3 3 3 3
1 2
⇒ E(X ) = ∫ x f(x) dx + ∫ x 2 f ( x )dx
2 2
0 1
1 2
= ∫ x .xdx + ∫ x 2 (2 − x ) dx
2
0 1
1 2
⎡ x 4 ⎤ ⎡ 2x 3 x 4 ⎤
= ⎢ ⎥ +⎢ − ⎥
⎣ 4 ⎦0 ⎣ 3 4 ⎦1
120
1 2 2 1
= + (8) − 4 − +
4 3 3 4
1 14 3 − 24 + 28 7
= −4+ = =
2 3 6 6
7 2 1
∴ Variance = −1 =
6 6
∞
MX(t) = E[etx] = ∫e
tx
Now f(x)dx
−∞
1 2
= ∫e tx
xdx + ∫ (2 − x )e tx dx
0 1
1 2
⎡ e tx ⎤ 1 e tx ⎡ e tx ⎤
2
e tx
= ⎢x ⎥ ∫
− (1) dx + ⎢ ( 2 − x ) ⎥ ∫
− ( − 1) dx
⎣ t ⎦0 0 t ⎣ t ⎦1 1 t
e t 1 tx
[ ] + ⎡⎢0 − 1. et ⎤⎥ + t1 [e ]
t
1 tx 2
= (1) − e 0 1
t t2 ⎣ ⎦
2
=
et 1 t
t t
[ et 1
]
− 2 e − 1 − + 2 e 2t − e t
t t
[ ]
1 t 1 e 2t e t
=− e + 2+ 2 − 2
t2 t t t
1 e 2 t 2e t 1 2t
= 2
+ 2 − 2 = 2 [e − 2et + 1]
t t t t
2
⎛ et −1⎞
MX(t) = ⎜⎜ ⎟
⎟
⎝ t ⎠
d ⎡ dr
μ1′ = M X ( t ) ⎢Θ μ ′r = r M X ( t )
Now dt ⎢ dt
t=0 ⎢ t=0
⎢
⎣
d
μ ′2 = M X (t)
dt
t=0
121
1 ⎡⎛ (2t ) 2 (2t ) 3 (2t ) 4 ⎞ ⎛ t2 t3 ⎞ ⎤
also MX(t) = ⎢⎜⎜1 + 2 t + + + + ..... ⎟⎟ − 2⎜⎜1 + t + + + ....⎟⎟ + 1⎥
t2 ⎣⎢⎝ 2! 3! 4! ⎠ ⎝ 2! 3! ⎠ ⎦⎥
1 ⎡ t2 t3 t4 ⎤
= ⎢ ( 4 − 2 ) + (8 − 2 ) + (16 − 2 ) + ......⎥
t2 ⎣ 2! 3! 4! ⎦
1 ⎡ 2t 2 6t 3 7 t 4 ⎤
= 2 ⎢
+ + + ......⎥
t ⎣ 2 6 12 ⎦
⎡ t ⎛ 7 ⎞ t2 ⎤
MX(t) = ⎢1 + + ⎜ ⎟ + ......⎥
⎣ 1! ⎝ 6 ⎠ 2! ⎦
t
μ1′ = coefficient of → Mean
1!
t2 7
μ ′2 = coefficient of =
2! 6
7 2 1
∴ Variance = μ ′2 − μ1′ 2 = −1 =
6 6
Proof: - By definition,
[
M X1+ X 2 +.....+ x n ( t ) = E e t ( X1+ X 2 +.....+ X n ) ]
= E[e tX1 tX 2
.e ........e tX n ]
If X1, X2,….., Xn are independent r.v., then function e tx1 , e tx 2 ,.....e tx n are also independent.
∴ [( ) ( )
M X1+ X 2 +.....+ x n ( t ) = E e tX1 .E e tX 2 ........E e tX n ( )]
= M X1 ( t ).M X 2 ( t )......M X n ( t )
122
Let us transform X to the new variable U by changing both the origin and scale in X as follows: -
Proof: - If X is a r.v., then
X−a
U= ⇒ X = a + hU where a & h are constants
h
Now MX(t) = E[etX] = E[et(a+hU)]
= E[eat, ethU]
= eat E[e(th)U]
where MX(t) is the m.g.f. of X about origin.
⇒ MX(t) = eatMU(th)
⇒ MU(th) = e−atMX(t)
t′
Put t h = t′ ⇒ t=
h
− at′
then MU ( t ′) = e ⎛ t ′ ⎞ = M.G.F of U (about origin)
X⎜ ⎟
h M
⎝h⎠
Standard normal variate
If X is a r.v., then the variable Z defined by
X−μ
Z= is called standard normal variate
σ
where μ, σ are Mean & standard derivation, respectively
Also E(Z) = 0
Variance (Z) = 1
⎛ X −μ ⎞ 1
Now E(Z) = E ⎜ ⎟ = E(X − μ )
⎝ σ ⎠ σ
1
E(Z) = [E(X) − μ ] = 1 [μ − μ ] = 0
σ σ
[ ]
2
⎛ X −μ ⎞ 1 2
and V(Z) = E ⎜ ⎟ = 2 E(X − μ )
⎝ σ ⎠ σ
1 2
⇒ V(Z) = .σ = 1 [Θ σ2 = E(X − μ)2]
σ2
If X ∼ N(μ, σ2),then Z ∼N(0, 1)
p ∞ tr r p t
( )∑ (qe )
∞
t r −1
⇒ MX(t) = ∑ e q = q qe
q r =1 r =1
p t
q
[
.qe 1 + qe t + (qe t ) 2 + ......
= ]
1
⇒ MX(t) = pet (1 − qe t ) [sum of G.P]
d (1 − qe t ).pe t − pe t .(−qe t )
M X (t)
Now μ1′ = dt = (1 − qe t ) 2
t=0 t=0
pe t − pqe 2 t + PQE 2T
= (1 − qe t ) 2
t=0
p
=
(1 − q ) 2
p p
⇒ μ1′ = 2
= 2 [Θ1 − q = p]
(1 − q ) p
1
⇒ μ 1′ = = Mean
p
d2
2
M X (t)
and μ ′2 = dt
t=0
d ⎛ pe t ⎞
⎜ ⎟
dt ⎜⎝ (1 − qe t ) 2 ⎟⎠
=
t=0
124
(1 − qe t )pe t − pe t .2(1 − qe t )(−qe t )
= (1 − qe t ) 4
t=0
(1 − qe t )pe t + 2pqe 2 t
= (1 − qe t ) 3
t=0
pe t + pqe 2 t
t 3
= (1 − qe )
t=0
p + pq p(1 + q ) 1 + q
= = = 2
(1 − q ) 3 p3 p
∴ Variance, μ2 = μ ′2 − μ 1′ 2
1+ q 1 q
⇒ μ2 = 2
− 2 = 2
p p p
The logarithm of m.g.f about origin of a r.v. X is called second m’g.f or cumulant generating
function (c.g.f.)
tr
i.e. KX(t) = log MX(t) = ∑Kr r!
r
tr
and Kr = coefficient of in KX(t)
r!
This Kr is known as rth cumulant
Now K(t) = log MX(t) = log[E(etx)]
⎡ ⎛ t 2X 2 trXr ⎞⎤
= log ⎢E⎜⎜1 + tX + + ....... + + ..... ⎟⎟⎥
⎢⎣ ⎝ 2! r! ⎠⎥⎦
⎡ ⎛ t2 t r ⎞⎤
= log ⎢1 + ⎜⎜ μ1′ t + μ ′2 + ...... + μ ′r ⎟⎟⎥
⎣⎢ ⎝ 2! r! ⎠⎥⎦
⎡⎛ t2 tr ⎞ 1⎛ t2 tr ⎞
⇒ KX(t) = ⎢⎜⎜ μ1′ t + μ ′2 + ....μ ′r + ...... ⎟⎟ − ⎜⎜ μ1′ + μ ′2 + .... + μ ′r + + ..... ⎟⎟
⎣⎢⎝ 2! r! ⎠ 2⎝ 2! r! ⎠
125
3
1⎛ t2 tr ⎞
+ ⎜⎜ μ1′ t + μ ′2 + .... + μ ′r + .... ⎟⎟
3⎝ 2! r! ⎠
1⎛ t2 tr ⎞
4 ⎤
− ⎜⎜ μ1′ t + μ ′2 + .... + μ ′r ⎟⎟ + .....⎥ ...(1)
4⎝ 2! r! ⎠ ⎥
⎦
t2 t3 t4
Now KX(t) = K1t + K2 + K 3 + K 4 + ...... …(2)
2! 3! 4!
Equating coefficients of like terms in equation (1), (2) we get
t
K1 = coefficient of = μ 1′ = Mean
1!
K2 = μ ′2 − μ1′ = μ 2 = (variance)
K3 = μ ′3 − 3μ ′2 μ1′ + 2μ1′3 = μ 3
( ) (
= μ ′4 − 4μ ′3μ1′ + 6μ ′2 μ1′ 2 − 3μ1′ 4 − 3 μ ′22 − 2μ ′2μ1′ 2 + μ1′ 4 )
⇒ K4 = μ4 − 3 K 22 = μ4 − 3 μ 22
dr
Also Kr = r K X ( t )
dt
Characteristic function
Characteristic function of a variable X is defined as φX(t) = E[eitX]
If X is continuous r. v., then
φX(t) = ∫ e itx f ( x ) dx
If X is discrete r. v, then
φX(t) = ∑ e itx f (x )
126
LESSSON 6 THEORETICAL DISCRETE DISTRIBUTIONS
BINOMIAL DISTRIBUTION
Let n represent the number of trials in an event.
x→ consecutive successes
(n−x)→ failure
then ppp…..p qq…..q = pxqn−x
But x successes in n trials can occur in nCx ways.
∴ Probability of x successes in any order in n trials = nCx px qn−x, here n and p are called
parameters.
Therefore, probability density function for a Binomial distribution is
P(X = x) = B(n, p; x)
= f(x) = nCx px qn−x, x = 0, 1, 2…., n
Also p+q=1 and ∑ f (x) = 1
n
⇒ μ ′r = ∑ x r nCx px qn−x
x =0
n
when r = 1, 1st moment, μ1′ = ∑ x nCx px qn−x
x =0
n x |n
⇒ μ1′ = ∑ px qn−x
x =0 |n − x |x
n
|n
= ∑ |n − x |x −1
px qn−x
x =1
n
|n −1
= np ∑ |n − x px−1 qn−x
x =1 | x −1
n
|n −1
= np ∑ | (n − 1) − (x − 1) | x −1
px−1 q(n−1)-(x−1)
x =1
127
n
= np ∑ n −1
C x −1p x −1 q ( n −1)−( x −1)
x =1
= np (q + p)n−1 [Θ ∑ f (x) = (q + p) n
]
=1
⇒ μ1′ = np = Mean
n n
= ∑ x ( x − 1) n C x p x q n − x + ∑ x n C x p x q n − x
x =0 x =0
n n
= ∑ x (x − 1) n C x p x q n −x + ∑ x n C x p x q n −x
x =2 x =1
n x ( x − 1) | n
= ∑ x ( x − 1) | x − 2 | n − x px qn−x + np
x =2
n n ( n − 1) | n − 2
⇒ μ ′2 = ∑ px qn−x+ np
x =2 |x − 2 |n − x
n
|n − 2
= n(n−1)p2 ∑ | x − 2 | (n − 2) − ( x − 2)
px−2 q(n − 2) − (x − 2) + np
x =2
n
= n(n−1)p2 ∑ n −2
C x −2 p x −2 q ( n −2)−( x −2) + np
x =2
= n(n−1)p2.(1) + np
⇒ μ ′2 = n(n−1)p2 + np
Also variance, μ2 = μ ′2 − μ 1′ 2
⇒ μ2 = n(n−1)p2 + np−n2p2
= n2p2 − np2 + np − n2p2
⇒ μ2 = np(1 − p) = npq
So, variance = npq
Also variance ≤ Mean
i.e. npq ≤ np as q ≤ 1
and variance = Mean if q = 1
128
n
Now μ ′3 = E (X 3 ) = ∑ x 3 f ( x )
x =0
n n
= ∑ x (x − 1)(x − 2) n C x p x q n−x + 3∑ x (x − 1) n C x p x q n −x
x =0 x =0
n
+ ∑ x n C x p x q n −x
x =0
n n
= ∑ x (x − 1)(x − 2) n C x p x q n −x + 3∑ x (x − 1) n C x p x q n−x
x =3 x =2
n
+ ∑ x n C x p x q n −x
x =1
n
|n
= ∑ x ( x − 1)( x − 2) x ( x − 1)( x − 2) | x − 3 | n − x p x q n−x
x =3
+ 3n ( n − 1) p 2 + np
n
|n − 3
= n(n−1)(n−2) ∑ p x q n − x + 3n (n − 1)p 2 + np
x =3 | x − 3 | (n − 3) − ( x − 3)
n
μ ′3 = n(n−1)(n−2)p3 ∑ n −3
C x −3 p x −3 q ( n −3)−( x −3) + 3n (n − 1)p 2 + np
x =3
⇒ μ ′3 = n(n−1)(n−2)p3 + 3n(n−1)p2 + np
129
= np(−q)(2p−p−q) = npq(q−p)
n
μ ′4 = ∑ x 4 n C x p x q n −x
x =0
⇒ μ4 = npq[1 + 3(n−2)pq]
Moment generating function (M.G.F) about origin for a Binomial distribution
MX(t) = E[etx] = ∑ e tx f (x )
n
= ∑ e tx n
C x p x q n −x
x =0
n
= ∑ n
C x q n − x (pe t ) x [Θetx = (et)x]
x =0
MX(t) = (q + pet)n
dr
μ ′r = M X (t)
Now dt r
t=0
d
μ1′ = M X (t)
⇒ dt
t=0
d
μ1′ = (q + pe t ) n
dt
t=0
t n −1 t
= n (q + pe ) pe
t=0
= np (q + p) n−1 = np
d2
μ ′2 = M X (t)
Also dt 2
t=0
d⎡
⎢
(
n q + pe t )
n −1
pe t ⎤
⎥⎦
= dt ⎣
t=0
130
= np
dt
[
d t
e (q + pe t ) n −1 ]
t=0
= np[et(q + pet)n−1 + 1 (n−1)(q + pet)n−2pet] t = 0
= np[ 1(q + p)n−1 + 1(n−1)(q+p)n−2p]
= np[1 + (n−1)p] = np+ n(n−1)p2
Pearson’s Coefficients
μ 32 μ4
β1 = , β2 = , α 1 = β1 , α 2 = β 2 − 3
μ 32 μ 22
(1 − 2p) 2 (1 − 6pq
β1 = , β2 = 3 +
npq npq
1 − 2p 1 − 6pq
α1 = , α2 =
npq npq
Characteristic function
φX(t) = E[eitx] = (q + peit)n
Cumulants
Cumulant generating function (c.g.f) of a Binomial distribution is
KX(t) = log MX(t)
= log (q + pet)n = n log (q + pet)
⎡ ⎛ t2 t3 ⎞⎤
= n log ⎢q + p⎜⎜1 + t + + + ...... ⎟⎟⎥
⎣⎢ ⎝ 2! 3! ⎠⎦⎥
⎡ ⎛ t2 t3 ⎞⎤
= n log ⎢(q + p) + p⎜⎜ t + + + ........ ⎟⎟⎥
⎣⎢ ⎝ 2! 3! ⎠⎦⎥
⎡ ⎛ t2 t3 ⎞⎤
= n log ⎢1 + p⎜⎜ t + + + ........ ⎟⎟⎥
⎣⎢ ⎝ 2! 3! ⎠⎦⎥
x2 x3
using log (1 + x) = x− + ......., we get
2 3
⎡ ⎛ t2 t3 ⎞ p2 ⎛ t 2 t3 ⎞
2
p3 ⎛ t 2 ⎞ ⎤
3
131
dr
r K X (t)
t dt r
Kr = coefficient of in KX(t) =
r! t=0
So K1 = np = μ1′
t2
K2 = np - np2 = coefficient of
2!
⇒ K2 = np(1−p) = npq = μ2 [Θp + q = 1]
K 3 np np 2 2 np 3
& = − +
3! 3! 2 2! 3
⇒ K3 = np−3np2 + 2p3n = np(1−3p+2p2)
= np(p−1)(2p−1)
= np(−q)[2p−p−q]
= npq(q−p) = μ3
Example: - p = probability of getting a head = ½
q = probability of not getting a head = ½
The probability of getting x heads in a random throw of 10 coins is
x 10− x 10
⎛ 10 ⎞ ⎛ 1 ⎞ ⎛ 1 ⎞ ⎛ 10 ⎞⎛ 1 ⎞
p(x) = ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ = ⎜ ⎟⎜ ⎟ ; x = 0, 1, 2, ......,10
⎝ x ⎠ ⎝2⎠ ⎝2⎠ ⎝ x ⎠⎝ 2 ⎠
∴ Probability of getting at least 7 heads is
P(X ≥ 7) = p(7) + p(8) + p(9) + p(10)
10
⎛ 1 ⎞ ⎧⎛ 10 ⎞ ⎛ 10 ⎞ ⎛ 10 ⎞ ⎛ 10 ⎞⎫
= ⎜ ⎟ ⎨⎜ ⎟ + ⎜ ⎟ + ⎜ ⎟ + ⎜ ⎟⎬
⎝ 2 ⎠ ⎩⎝ 7 ⎠ ⎝ 8 ⎠ ⎝ 9 ⎠ ⎝ 10 ⎠⎭
120 + 45 + 10 + 1 176
= =
1024 1024
132
Recurrence Relation
d
μr+1 = pq[nr μr−1 + μr]
dp
where μr is the rth moment about mean.
n
Now, μr = E[(X−np)r] = ∑ (x − np) r n
C x p x q n −x
x =0
[ ]
n
+ ∑ ( x − np) r n C x q n − x xp x −1 + p x (n − x )q n − x −1 (−1)
x =0
⎡ dq ⎤
⎢Θ p = 1 − q, q = 1 − p ∴ dp = −1⎥
⎣ ⎦
n n
⎡x n − x⎤
= −nr ∑ ( x − np) r −1 f ( x ) + ∑ ( x − np) r n C x p x q n −x ⎢ −
x =0 x =0 ⎣p q ⎥⎦
n
( x − np)
= −nr μr−1 + ∑ (x − np) r n C x p x q n−x pq
x =0
⎡ x (n − x ) xq − np + xp x (p + q ) − np x − np ⎤
⎢∴ p − q = pq
=
pq
=
pq ⎥⎦
⎣
dμ r n
( x − np) r +1 n
⇒ = −nrμ r −1 + ∑ C x p x q n −x
dp x =0 pq
1
= − nrμr−1 + μ r +1
pq
133
M.G.F about its Mean
MX−np(t) = [qe−pt + petq]n
n
Now by definition MX−np(t) = ∑ e ( x − np) t f ( x )
x =0
n n x n−x
⇒ MX−np(t) = ∑ e ( x − np) t C x p q
x =0
n
= ∑ e xt e −npt n
C x p x q n −x
x =0
n
= e−npt ∑ n C x q n −x (pe t ) x
x =0
= e−npt(q + pet)n
= (e−pt)n(q + pet)n
= [qe−pt + pet(1−p)]n
⇒ MX−np(t) = [qe−pt + peqt]n [Θ p + q = 1]
⎡ ⎛ p 2 t 2 p3t 3 ⎞ ⎛ q 2 t 2 q3t 3 ⎞ ⎤
n
⎡ t2 t3 ⎤
= 1 + n ⎢pq + pq (q − p) + .......⎥
⎣ 2! 3! ⎦
2
n (n − 1) ⎡ t 2 t3 ⎤
+ ⎢ pq + pq ( q − p ) + .......⎥ + ......
2! ⎣ 2! 3! ⎦
t2 t3
= 1 + npq + npq (q − p) + .......
2! 3!
∴ μ1 = 0
134
t2
μ2 = coefficient of = npq
2!
t3
μ3 = coefficient of = npq(q − p)
3!
Example: - In 8 throw of a dice, 5 or 6 is considered as success. Find the mean number of
successes and S.D.
2 1 2
Solution: - Here n = 8, p = = ,q=
6 3 3
1 8
∴ Mean = np = 8 × =
3 3
8 2 4
and SD = npq = × =
3 3 3
Example: - 6 dice are thrown 729 times. How many times due you expect at least 3 dice to show
five or six?
Solution: - n = 6, N = 729
2 1 2
p= = , x =3, q =
6 3 3
∴ P(X≥3) = f(3) + f(4) + f(5) +f(6) …(1)
Now f(x) = nCx px qn−x
6
∴ (1) ⇒ P(X≥3) = ∑ n C x p x q n −x
x =3
3 3 4 2 5 6
⎛1⎞ ⎛ 2⎞ ⎛1⎞ ⎛ 2⎞ ⎛1⎞ ⎛ 2⎞ ⎛1⎞ 233
= C 3 ⎜ ⎟ ⎜ ⎟ + 6 C 4 ⎜ ⎟ ⎜ ⎟ + 6⎜ ⎟ ⎜ ⎟ + ⎜ ⎟ =
6
135
n
F(x) = N f(x) = N ∑ n C x p x q n −x
x =0
for x = 0
F(0) = Nqn …(1)
f (x) n
C x p x q n −x
Also = n
f ( x − 1) C x −1 p x −1q n − x +1
Nf ( x ) n − x +1 p
= .
Nf ( x − 1) x q
F( x ) ⎛ n − x +1⎞ p
⇒ =⎜ ⎟
F( x − 1) ⎝ x ⎠q
⎛ n − x +1⎞ p
⇒ F(x) = ⎜ ⎟. F(x−1) …(2)
⎝ x ⎠q
where x = 1, 2, …..,n
POISSON DISTRIBUTION
Poisson distribution can be obtained from Binomial distribution under three conditions
(i) no. of trials is large i.e. n → ∞
(ii) probability of success is very small i.e. p → 0
(iii) the mean no. of successes is finite, i.e. np = m(say)
In case of Binomial distribution,
f(x) = n C x p x q n − x , x = 0, 1, 2,….n
Taking limit case of f(x) in Binomial distribution, we get f(x) for Poisson distribution,
m x e −m
f(x) = , x = 0, 1, 2, 3,…..
x!
Now n → ∞, p → 0, np = m
m m
i.e. p= , q = 1− as n → ∞
n n
Probability of x successes in Binomial distance is
f(x) = n
C x p x q n−x
Probability of x successes in Poisson distance is
136
f(x) = lim n C x p x q n − x
n →∞
x n−x
⎛m⎞ ⎛ m⎞
= lim n C x ⎜ ⎟ ⎜1 − ⎟
n →∞ ⎝n⎠ ⎝ n⎠
x n−x
n! ⎛m⎞ ⎛ m⎞
= lim ⎜ ⎟ ⎜1 − ⎟
n →∞ x!( n − x )! ⎝ n ⎠ ⎝ n⎠
⎛ m 2 m3 ⎞
= lim ⎜⎜1 − m + − + ..... ⎟⎟
n →∞
⎝ 2! 3! ⎠
= e−m
∴ from (1)
mx
f(x) = e −m , x = 0, 1, 2,…..
|x
137
∞
∴ μ1′ = E[X] = ∑ x f (x)
x =0
∞
mx ∞
xe − m m x
= ∑ x e −m |x
=∑
x| x − 1
x =0 x =0
∞
e −m m x
= ∑ | x −1
x =1
∞
m x −1
= me −m ∑ | x −1
x =1
⎡ m m2 ⎤
= me −m ⎢1 + + + .....⎥
⎣⎢ |1 | 2 ⎦⎥
⇒ μ1′ = me−m em = m
⇒ μ1′ = m = Mean
∞
mx
Now μ ′2 = E(X2) = ∑ x 2 e −m
|x
x =0
∞ ⎡ x ( x − 1) + x ⎤ −m x
= ∑⎢ ⎥e m
x =0 ⎢
⎣ | x ⎥⎦
⎡ ∞ x ( x − 1) x ∞ x m x ⎤
= e−m ⎢ ∑ m +∑ ⎥
⎣⎢ x =0 | x ⎥
x =0 | x ⎦
∞ ∞
mx mx
= e−m ∑ | x − 2 + e ∑
−m
x =2 x =1 | x −1
∞
m x −2
=m e 2 −m ∑ |x − 2 + m
x =2
= m2 e−m.em + m
⇒ μ ′2 = m2 + m
∴ Variance, μ2 = μ ′2 − μ1′
⇒ μ2 = m2 + m − m2 = m ⇒ μ2 = m
So mean = Variance
138
Moment generating function about origin
M.G.F about origin is by definition
∞
MX(t) = E[etx] = ∑ e tx f (x)
x =0
∞
e −m m x ∞
(me t ) x
= ∑e tx
|x
=e −m
∑ |x
x =0 x =0
⎡ t (me t ) 2 ⎤
−m ⎢1 + me + + .....⎥
=e
⎢⎣ |2 ⎥⎦
t
= e−m. e me
t −1)
⇒ MX(t) = e m ( e ...(1)
∞
e−m m x
= ∑e t(x−m )
|x
x =0
∞
mx
= ∑ e tx e −mt e −m |x
x =0
∞
(me t ) x
= e−(1+t)m ∑ |x
x =0
t t − t −1)
= e−m(1+t) e me = e m ( e …(2)
⎛ t 2 t3 ⎞ ⎛ t 2 t3 ⎞
m ⎜ 1+ t + + +..... ⎟ − t −1 m ⎜ + +..... ⎟
⎜ |2 |3 ⎟ ⎜ 2 |3 ⎟
⎝ ⎠ ⎝| ⎠
=e =e
dr
r
M X (t)
Now μ ′r = dt
t=0
139
d
M X (t)
∴ μ1′ = dt
t=0
d m ( et −1)
e
= dt
t=0
m(e t −1)
=e me t
t=0
μ1′ = m
d2
2
M X (t)
and μ ′2 = dt
t=0
d ⎡ t m ( e t −1) ⎤
me e
dt ⎢⎣ ⎥⎦
=
t=0
m ⎡e t m ( e t −1)
me t ⎤
t −1)
+ e t e m(e
⎢ ⎥⎦
= ⎣
t=0
= m[1.e0 + 1. 1. m. 1]
= m(1 + m) = m2 + m
Characteristic function
φX(t) = e m( e
it −1)
[ ]
= E e itx
tr
Kr = coefficient of in KX(t)
|r
∴ K1 = m, K2 = m, K3 = m, ……
i.e. Kr = m , for r =1, 2, 3,…..
All cumulants are equal for Poisson distribution
140
Recurrence Relation for Moments
dμ r
μr+1 = m[rμ r −1 + ]
dm
Now rth central moment by definition is
μr = E[(X − m)r]
∞ ∞
e −m m x
= ∑ ( x − m) r f ( x ) = ∑ ( x − m) r |x
x =0 x =0
∞ ⎡
dμ r mx ⎤ ∞ ⎡ mx ⎤
∴ = ∑ ⎢r ( x = m) r −1 (−1)e −m ⎥ ∑⎢
+ ( x − m ) r −m
e ( −1) ⎥
dm x =0 ⎢⎣ | x ⎥⎦ x =0 ⎢⎣ | x ⎥⎦
∞
m x −1
+ ∑ ( x − m) r e − m x
x =0 |x
∞
e −m m x ∞
e −m m x ⎡ x ⎤
= −r ∑ ( x − m) r −1 + ∑ ( x − m) r ⎢⎣ m − 1⎥⎦
x =0 |x x =0 |x
∞
e −m m x ∞
e −m m x ⎛ x − m ⎞
= −r ∑ ( x − m) r −1
|x
+ ∑ ( x − m) r ⎜
|x ⎝ m ⎠
⎟
x =0 x =0
−m x
1 ∞ r +1 e m
= −r μr−1 + ∑ ( x − m)
m x =0 |x
dμ r 1
⇒ = − rμ r −1 + μ r +1
dm m
dμ r 1
⇒ + rμ r −1 = μ r +1
dm m
dμ r
⇒ μr+1 = m[rμ r −1 + ]
dm
Recurrence relation or Fitting of Poisson distribution
e −m m x
f(x) =
|x
e − m m x +1
f(x + 1) =
|x +1
f ( x + 1) m
∴ =
f (x) x +1
141
mf ( x )
⇒ f(x + 1) = …(1)
x +1
Now Nf(x) = F(x)
∴ F(0) = Nf(0) = Ne−m
⎛ m ⎞
∴ (1) ⇒ Nf(x + 1) = ⎜ ⎟, f ( x ) N
⎝ x +1⎠
m
⇒ F(x + 1) = F( x ) , x = 0,1,2,..... …(2)
x +1
Put x = 0 in (2), w get
F(1) = m F(0) = m Ne−m
m m2
F(2) = F(1) = Ne −m
2 2
m m3
F(3) = F( 2 ) = Ne −m ,.....
3 6
Additive Property of Poisson variate
If X1, X2 are two independent Poisson variates with parameters m1, m2.Then their sum X1 + X2 is
also a Poisson varite with Parameter (m1 + m2)
i.e. X1 + X2 ∼ P(m1 + m2;x)
Proof: - If X is a Poisson variable, then
e −m m x
f(x) = , x = 0, 1, 2….
|x
t −1)
& M x1 ( t ) = e m1 ( e
t −1)
M X2 ( t ) = e m2 (e
“The MGF of sum of two independent r.v. is equal to the product of their MGF.”
By using this result, we have
M X1+ X 2 ( t ) = M X1 ( t ) . M X 2 ( t )
t −1) t −1)
= e m1( e . e m2 ( e
t −1)( m + m )
= e (e 1 2
142
2
Example: - Show that in a Poisson distribution with unit mean, deviation about mean is times
e
the standard deviation.
Solution: - Here mean = 1
Variance = SD = 1 [ΘIn Poisson distance mean = variance]
∴ M.D = Mean Deviation = E[|X−1|]
∞
| x − 1 | e −1
= ∑ |x
x −0
∞
| x −1|
=e −1 ∑ |x
x =0
1⎡ 1 2 3 ⎤
= ⎢1 + + + + .....⎥
e ⎢⎣ | 2 |3 | 4 ⎥⎦
Now general term of the series
1 2 3
+ + + .....is
|2 |3 |4
n n +1−1
General term = =
|n +1 |n +1
1 1
= −
|n |n +1
Putting n = 1, 2, 3….
1 ⎡ ⎛⎜ 1 ⎞ ⎛1 1⎞ ⎛1 1⎞ ⎤
⎟ + ⎜ − ⎟ + ⎜ − ⎟ + .....⎥
M.D = ⎢1 + ⎜1 − ⎟ ⎜ | 2 |3 ⎟ ⎜ |3 | 4 ⎟
e ⎢⎣ ⎝ |2 ⎠ ⎝ ⎠ ⎝ ⎠ ⎥⎦
1 2 2
= (1 + 1) = × 1 = × S.D.
e e e
143
⎡ ⎛ pe t ⎞⎤
∴ log MX−np(t) = −npt + n log ⎢q ⎜⎜1 + ⎟⎥
⎟
⎣⎢ ⎝ q ⎠⎦⎥
⎛ pe t ⎞
= −npt + n log q + n log ⎜⎜1 + ⎟
⎟
⎝ q ⎠
Put np = m, we get
⎡ ⎤
⎢ t ⎥
⎛ m⎞ me
Log MX−np(t) = −mt + n log ⎜1 − ⎟ +n log ⎢1 + ⎥
⎝ n⎠ ⎢ ⎛ m ⎞⎥
⎢ n ⎜⎝1 − n ⎟⎠ ⎥
⎣ ⎦
⎡ ⎧ ⎫
2
⎢
⎡ m m 2 m3 ⎤ me t 1 ⎪⎪ me t ⎪⎪
= −mt+n ⎢− − 2 − 3 ....⎥ + n ⎢ − ⎨
⎢ ⎛ m ⎞ 2 ⎛ m ⎞⎬
⎣ n 2n 3n ⎦ ⎢ n ⎜1 − ⎟ ⎪ n ⎜1 − ⎟ ⎪
⎝ n ⎠ ⎪⎩ ⎝ n ⎠ ⎪⎭
⎣⎢
3⎤
⎧ ⎫
⎪ t ⎪⎪ ⎥
1 ⎪ me ⎥
+ ⎨ ⎬
3 ⎪ ⎛ m ⎞⎪ ⎥
n ⎜1 − ⎟ ⎥
⎪⎩ ⎝ n ⎠ ⎪⎭ ⎥⎦
2
⎧ ⎫
me ⎪
1 me t ⎪
t
m2 ⎛ 1 ⎞
= −mt − m + − ⎨ m⎬ − + 0⎜ 2 ⎟
⎛ m ⎞ 2n ⎪1 − ⎪ 2n ⎝n ⎠
⎜1 − ⎟
⎝ n⎠ ⎩ n⎭
me t ⎛1⎞
= −mt −m + + 0⎜ ⎟
m ⎝n⎠
1−
n
As n→∞, we have
lim log M X −np ( t ) = − mt − m + me t
n →∞
t − mt −m
⇒ lim M X−np ( t ) = e me
n →∞
t − t −1)
= e m(e
= M.G.F about mean of Poisson distance
144
Example: - If X is a Poisson variate s.t. P(X = 1) = P(X = 2), find P(X = 4).
Solution: - Probability of x successes in Poisson distribution is
e −m m x
P(X = x) = f(x) =
|x
e −m m
∴ P(X = 1) = = e−m m …(1)
|1
e −m m 2 m 2 −m
P(X = 2) = = e …(2)
|2 2
m 2 −m
m e−m = e ⇒m=2
2
Therefore for X = 4
e −m m 4 e −2 (2) 4
P(X = 4) = =
|4 24
16 −2
= e = 0.09
24
Let e−2 = a
⇒ log10 a = −2 log10 e
⇒ log10 a = −2(0.4343) = −0.8686
= T.1314
⇒ a = antilog (T.1314) = 0.1353
Example: -Six coins are tossed, 6400 times, using Poisson distribution, what is the probability of
getting six heads x times?
1
Solution: - When a coin is tossed, probability of getting head is =
2
2
⎛1⎞ 1
When 6 coins are tossed, probability of getting 6 heads = ⎜ ⎟ = = p(say)
⎝2⎠ 64
1
Mean = np = 6400 × = 100 (Mean)
64
145
⇒ m = 100
Probability of getting six heads x times
e −m m x
= with m = 100
|x
e −100 (100) x
=
|x
Example: - After correcting 50 pages of the proof of a book, the proof reader finds that there are
on the average 2 errors of 5 pages. How many pages would one expect to find with 0, 1, 2, 3 & 4
errors in 1000 pages of first print of the book?
2
Solution: - Now Mean = = 0.4 = M (say)
5
X→ Poisson distance,
e −m m x
f(x) = P(X = x) =
|x
e −0.4 (0.4) x
=
|x
Expected number of pages with 1000 pages is
146
3 0. 4 7.1298 ≈ 7
f(3) = f (2) = 0.0071298
2 +1
4 0 .4 ≈1
f(4) = f(3) = 0.0007129
3 +1
1
Example: - In a certain factory making blades, there is a small chance for any blade to be
500
defective. The blades are supplied in packets of 10. Use Poisson distance to calculate the
approximate. Number of packets containing no defective, one defective & 2 defective blades
respectively in a consignment of 10000 packets.
(e−0.02 = 0.9802)
1
Solution: - Let p→probability of blade being defective =
500
n = 10
1
∴ Mean = m = np = = 0.02
50
Number of packets containing x defective blades,
Ne − m m x
F(x) =
|x
10000 × 0.9802(0.02) x
=
|x
9802(0.02) x
F(x) =
|x
X F(x)
0 9802(0.02) 0
F(0) = = 9802
|0
1 9802(0.02)1
F(1) = =196
|1
147
2 9802(0.02) 2
F(2) =
|2
Mode of Poisson distribution
Let the most probable values of r.v. X be r i.e. r is mode.
Then P(X = r) should be greater than P(X = r + 1) and P(X = r− 1)
i.e
e − m m r −1 e −m m r e − m m r +1
< >
| r −1 |r |r +1
m r −1 m.m r −1 m r −1m 2
⇒ < >
|r −1 r | r − 1 (r + 1)r | r − 1
m m2
⇒ 1< >
r r (r + 1)
⇒ r(r + 1) < m(r + 1) > m2
⇒ r < m & (r + 1) > m
⇒ r<m &r>m−1
⇒ m−1<r<m
Thus if m is an integer, there shall be two modes m and (m−1).
If m is not an integer, then mode is the integral value between (m−1) & m.
Example: - A Poisson distance a double mode at X = 3 and X = 4, what is the probability that X
will have one or the other of these two values.
e −m m x
Solution: - P(X = x) =
|x
e −m m 3 ⎫
∴ P(X = 3) = ⎪
|3 ⎪
⎬ …(1)
e −m m 4 ⎪
& P(X = 4) =
| 4 ⎪⎭
Since the Poisson distance has a double mode at X = 3 and X = 4, so
P(X = 3) = P(X = 4)
e −m m3 e −m m 4
⇒ =
|3 |4
148
|4
⇒ m= =4
|3
e −4 4 3 32 −4
∴ from (1), P(X = 3) = = e
|3 3
e −4 4 4 e −4 .4.4 3
& P(X = 4) = =
|4 4 |3
e − 4 .4 3 32 −4
= = e
|3 3
⎛X−m⎞ 1
then E(Z) = E ⎜ ⎟= E (X − m)
⎝ m ⎠ m
1
= [E(X) − E(m)]
m
1
= [m − m]= 0
m
2
⎛X−m⎞
and variance = E ⎜ ⎟
⎝ m ⎠
1 1
= E(X − m)2 = μ2
m m
149
1
= .m = 1
m
1 2
t
Also find M.G.F of this variable and show that it approaches e2 as m→∞
X−m
Let Y =
m
4 ⎡ t ( X −m ) ⎤
then MY(t) = E[ e t ] = E ⎢e m ⎥
⎢⎣ ⎥⎦
−
mt ⎛ tx ⎞
=e m E ⎜e m ⎟
⎜ ⎟
⎝ ⎠
tx
∞
e −m m x
=e − mt
∑ |x
e m
x =0
∞
(me t / m x
)
= e− mt
e −m ∑
x =0 |x
−m− m t
⎡ me t / m
(me t / m 2
) ⎤
=e ⎢1 + + + .....⎥
⎣⎢ |1 |2 ⎦⎥
t/ m
= e −m− mt
e me
⎛ t t2 t3 ⎞
−m− m t + m ⎜ 1 + + + +.... ⎟
⎜ m |2 |3 m m ⎟
⎝ ⎠
=e
t2 t3
+ +.....
2 |3 m
MY(t) = e
2/2
As m→∞, lim M 4 ( t ) = e t which is MGF of standard normal variate (Normal Distribution)?
m →∞
150
GEOMETRIC DISTRIBUTION
Suppose we have a series of independent trials or repetitions and on each trial or repetition, the
probability of success ‘p’ remains the same. Then the probability preceding the first success is
given by
pqx
where X→Number of failures preceding the 1st success in series of independent trials.
and p = P(success in a single trial)
eg., in tossing a coin, the possible outcomes of experiment are
S, FS, FFS, ……FF,…..FS,……..
X = 0, 1, 2,……, x
and corresponding probability are
p, qp, q2p,…….., qxp…
∴ Probability distribution of X,
⎧pq x , x = 0,1,2,......∞
f(x) = P(X = x) = ⎨
⎩0 , otherwise
Also ∑ f (x) = 1
x
∞ ∞
Now ∑ f (x) = ∑ pq x
x =0 x =0
1
= p. =1
p
Moments of Geometric distribution
∞ ∞
μ1′ = E (X) = ∑ xf ( x ) =∑ xpq x
x =0 x =1
⎛∞ ⎞
= pq ⎜⎜ ∑ xq x −1 ⎟⎟
⎝ x =1 ⎠
= pq(1 + 2q + 3q2 + ……)
151
= pq(1 − q)−2
1 q
⇒ μ 1′ = pq. = …(1)
p2 p
q
⇒ μ1′ = Mean =
p
∞ ∞
μ ′2 = ∑ x 2 f ( x ) = ∑ [x ( x − 1) + x ]pq x
x =0 x =0
∞ ∞
= ∑ x(x − 1) pq x + ∑ x pq x
x =2 x =1
∞
x ( x − 1) x −2 q
= 2pq2 ∑ 2.1
q +
p
[from (1)]
x =2
q 2q 2 q
= 2pq2(1−q)−3 + = +
p p2 p
2q 2 q q 2 q 2 q
= + − = +
p2 p p2 p2 p
q q
= 2
(q + p) = 2
p p
∞
= p ∑ (qe t ) x = p[1 − qe t ]−1
x =0
p
=
1 − qe t
152
pq pq q
= 2
= 2 =
(1 − q ) p p
d2 d ⎡ pqe t ⎤
M X (t) ⎢ ⎥
dt 2 dt ⎣ (1 − qe t ) ⎦
′
μ2 = =
t=0 t=0
⎡ (1 − qe t ) 2 e t − e t .2(1 − qe t )(−qe t ) ⎤
⎢ ⎥
(1 − qe t ) 4
= pq ⎣ ⎦
t=0
⎡ (1 − qe t )e t + 2qe 2 t ⎤
⎢ ⎥
⎣ (1 − qe t ) 3 ⎦
= pq
t=0
⎡ (1 − q ) + 2q ⎤
= pq ⎢ 3 ⎥
⎣ (1 − q ) ⎦
⎛ p + 2q ⎞ q 2q 2
= pq ⎜⎜ 3 ⎟⎟ = + 2
⎝ p ⎠ p p
d3
3
M X (t)
μ ′3 = dt
t=0
d t
[
e (1 − qe t ) −2 + 2qe 2 t (1 − qe t ) −3 ]
= pq dt
t=0
= pq [et (1−qet)−2 + et(−2) (1−qet)−3 (−qet) + 2qe2t. 2(1−qet)−3 + 2qe2t(−3)
(1−qet)−4 (−qet)]
= pq[(1−q)−2 + 2q(1−q)−3 + 4q(1−q)−3 + 6q2e(1−q)−4]
⎡ 1 6q 6q 2 ⎤
⇒ μ ′3 = ⎢ 2 + 3 + 4 ⎥ pq
⎣p p p ⎦
q 6q 2 6q 3
μ ′3 = + + 3
p p2 p
q 6q 2
= + (p + q)
p p3
q 6q 2 q 6q 2
= + 3 .1 = + 3
p p p p
153
d4
4
M X (t)
μ ′4 = dt
t=0
= pq
d t
dt
[
e (1 − qe t ) −2 + 2qe 2 t (1 − qe t ) −3 + 4qe 2 t (1 − qe t ) −3
+ 6q 2 e 3t (1 − qe t ) −4 ]
t=0
[
= pq e t (1 − qe t ) −2 + e t qe t .2(1 − qe t ) −3 + 4qe 2 t (1 − qe t ) −3
+ 6q 2 e 3t (1 − qe t ) −4 + 8qe 2 t (1 − qe t ) −3 + 12q 2 e 3t (1 − qe t ) −4
154
LESSON 7 THEORETICAL CONTINUOUS DISTRIBUTIONS
Uniform or rectangular distribution
A continuous r.v. X is said to have continuous uniform (or rectangular) distribution over an
interval [a, b] if its p.d.f. is given by
⎧K , a≤x≤b
f(x) = ⎨
⎩0 , otherwise
b
also ∫ f ( x ) dx = 1
a
b
⇒ ∫ K dx = 1 ⇒ [Kx ]ab = 1
a
1
⇒ K(b−a) = 1 ⇒K=
b−a
⎧ 1
⎪ , a≤x≤b
∴ f(x) = ⎨ b − a
⎪⎩0 , otherwise
f(x)
1
k=
b−a
X
a b
Moments
[ ]
b
μ ′r = E X r = ∫ x r f ( x ) dx
a
b
⎛ 1 ⎞
∴ μ ′r = ∫ x r ⎜ ⎟ dx
a ⎝ b − a ⎠
b
1 ⎛ x r +1 ⎞
= ⎜ ⎟
(b − a ) ⎜⎝ r + 1 ⎟⎠ a
155
1 ⎡b r +1 − a r +1 ⎤
=
(b − a )(r + 1) ⎢⎣ ⎥⎦
1 ⎛ b2 − a 2 ⎞ b + a
∴ μ1′ = ⎜ ⎟=
(b − a ) ⎜⎝ 2 ⎟⎠ 2
b+a
⇒ μ 1′ = Mean =
2
1 (b − a )(b 2 + a 2 + ab)
μ ′2 = (b 3 − a 3 ) =
(b − a ).3 3(b − a )
1 2
= (b + a2 + ab)
3
Variance, μ2 = μ ′2 − μ1′ 2
1 2 1
= (b + a2 + ab) − (b2 + a2 + 2ab)
3 4
b 2 a 2 2ab
= + −
12 12 12
1 2 (a − b) 2
= (b + a2 + 2ab) =
12 12
1
⇒ μ2 = (b−a)2
12
=
1
t (b − a )
[e bt − e at ]
Characteristic function
b
e ibt − e iat
φX(t) = ∫ e itx f ( x ) dx =
a
it (b − a )
156
Mean deviation
b
(a + b ) 1
E[|X−Mean|] = ∫ x−
2
.
(b − a )
dx
a
b
1 ⎛a+b⎞
= ∫
(b − a ) a
x −⎜ ⎟ dx
⎝ 2 ⎠
…(1)
a+b
Put t = x− ⇒ dt = dx
2
(a + b ) a − b
when x = b, t = a − =
2 2
(a + b ) b − a
when x = b, t = b − =
2 2
b −a
2
1
∴ (1) ⇒ E[|X − Mean|] =
(b − a ) ∫ | t | dt
⎛ b −a ⎞
−⎜ ⎟
⎝ 2 ⎠
b −a
b −a / 2
2 2 ⎡t2 ⎤ 2
=
(b − a ) ∫ t dt = ⎢ ⎥
(b − a ) ⎣ 2 ⎦ 0
0
2 (b − a ) 2 b − a
= . =
2(b − a ) 4 4
1
Example: - Show that for the rectangular distance if f(x) = , −a < x < a,
2a
sinh at
The MGF about origin is .
at
a 2n
Also show that moments of even order are given by μ2n =
2n + 1
Solution: - By definition,
[ ]
a
1 1 tx a
∫e
tx
MX(t) = . dx = e −a
−a
2a 2at
1 at 1
= (e − e − at ) = .2 sinh at
2at 2at
157
sinh at
⇒ MX(t) =
at
1⎡ (at ) 3 (at ) 5 ⎤
= ⎢at + + + ......⎥
at ⎣ 3! 5! ⎦
(at ) 2 (at ) 4
=1+ + + ....... …(1)
3! 5!
Since there are no terms with odd powers of t in MX(t), all moments of odd order about origin are
zero i.e.
μ ′2 n +1 (about origin) = 0
In particular, μ1′ = 0 i.e. Mean = 0
[Θ μ ′r = E[X r ], μ r = E[(X − μ ) r ]
t 2n
μ ′2n = μ 2n = Coefficient of in MX(t)
| 2n
a 2n
⇒ μ2n = [using (1)]
2n + 1
f(x)
x 158
The Cumulative Distribution function
F(x) is given by,
x x
F(x) = ∫ f (x ) dx = θ ∫ e −θx dx
0 0
⎧⎪ − θx
= ⎨1 − e , x≥0
⎪⎩0 , otherwise
F(x)
1
F(x)
∞
= θ ∫ e (θ− t ) x dx
0
∞
⎡ e −( θ − t ) x ⎤
= θ⎢ ⎥
⎣ − (θ − t ) ⎦ 0
θ
= , θ>t
θ−t
θ θ 1
⇒ MX(t) = = =
θ−t ⎛ t⎞ t
θ⎜1 − ⎟ 1 −
⎝ θ⎠ θ
−1
⎛ t⎞ t t2 t3
= ⎜1 − ⎟ = 1+ + + + ....... …(1)
⎝ θ⎠ θ θ θ
∞ r
⎛t⎞
= ∑ ⎜ ⎟
⎝θ⎠
r =0
159
t | 2 t 2 |3 t 3
(1)⇒ MX(t) = 1 + + + + .......
θ θ 2 | 2 θ3 |3
tr
∴ μ ′r = E[Xr] = coefficient of in MX(t)
|r
|r
= , r = 1, 2, 3,........
θ2
1
∴ Mean = μ 1′ =
θ
2
μ ′2 =
θ2
2 1 1
Variance = μ2 = μ ′2 − μ1′ 2 = 2
− 2 = 2
θ θ θ
Also by definition,
∞ ∞
Mean = ∫ x f ( x ) dx = ∫ x θ e −θx dx
0 0
⎡⎛ e −θx ⎞
∞ ∞
e − θx ⎤
= θ ⎢⎜⎜ x ⎟ − ∫ (1) dx ⎥
⎢⎝ − θ ⎟ − θ ⎥
⎣ ⎠0 0 ⎦
∞
⎡ e − θx ⎤
= θ ⎢0 + (−1) 2 ⎥
⎣ θ ⎦0
= −
θ
e [ ]
1 −θx ∞
0 =
−1
θ
[0−1] =
1
θ
∞ ∞
E(X ) = μ ′2 = ∫ x f ( x ) dx = ∫ x 2 .θe −θx dx
2 2
0 0
⎡⎛ e −θx ⎞
∞ ∞
e −θx ⎤
= θ ⎢⎜⎜ x 2 . ⎟ − ∫ 2 x. dx ⎥
⎢⎝ −θ ⎟ − θ ⎥
⎣ ⎠0 0 ⎦
⎡ 2
∞ ⎤
= θ ⎢0 − 0 + ∫ xe −θx dx ⎥
⎢⎣ θ0 ⎥⎦
2⎡ ⎤ 2 1 2
∞
⎢θ ∫ xe
−θx
= dx ⎥ = . = 2
θ ⎣⎢ 0 ⎥⎦ θ θ θ
160
Variance = E[X−E(x)]2
2
⎡ 1⎤ 1
= E ⎢X − ⎥ = E(X 2 ) − 2
⎣ θ⎦ θ
2 1 1
= 2
− 2 = 2
θ θ θ
Mean Deviation
⎡ 1 ⎤
E|X−E(X)| = E ⎢ X − ⎥
⎣ θ ⎦
∞
1 −θx
= ∫ x−
θ
θe dx
0
∞ ∞
θ
= ∫ | xθ − 1 | e −θx dx = ∫ | xθ − 1 | e −θx dx
θ0 0
Put θx = y ⇒ θdx = dy
when x = 0 ⇒ y = 0
x=∞⇒y=∞
∞
dy
∴ E[|X−E(X)|] = ∫ | y − 1 | e − y
0
θ
∞
1
= ∫ | y − 1 | e − y dy …(1)
θ0
1⎡ ⎤
1 ∞
∴(1) E[|X−E(X)|] = ⎢ ∫ (1 − y)e dy + ∫ ( y − 1) − y dy⎥
−y
θ ⎣⎢ 0 1 ⎥⎦
1 ⎡⎛
11 ∞
e −y ⎞ e −y ⎛ e −y ⎞
= ⎢⎜⎜ (1 − y) ⎟ − ∫ (−1) dy + ⎜ ( y − 1) ⎟
θ ⎢⎝ − 1 ⎟⎠ 0 0 −1 ⎜
⎝ − 1 ⎟⎠1
⎣
161
e−y ⎤
∞
− ∫ 1. dy ⎥
1
− 1 ⎦⎥
1⎡
1 ∞⎤
⎛ e −y ⎞ ⎛ e −y ⎞
= ⎢− 0 + 1 − ⎜⎜ ⎟ +0−0+⎜ ⎟ ⎥
θ⎢ ⎝ − 1 ⎟⎠ 0 ⎜ −1 ⎟
⎝ ⎠1 ⎥
⎣ ⎦
=
1
θ
[
1 + (e −1 − 1) − (0 − e −1 ) ]
1 −1 2
= [e + e−1] = e−1
θ θ
Cumulants
The c.g.f. is
−1
⎛ t⎞ ⎛ t⎞
KX(t) = log MX(t) = log ⎜1 − ⎟ = − log⎜1 − ⎟
⎝ θ⎠ ⎝ θ⎠
⎡ t 1 ⎛ t ⎞ 2 1 ⎛ t ⎞3 ⎤
= ⎢ + ⎜ ⎟ + ⎜ ⎟ + .......⎥
⎢⎣ θ 2 ⎝ θ ⎠ 3 ⎝ θ ⎠ ⎥⎦
r
∞
⎛t⎞ 1 ∞
t r ⎛ |r −1 ⎞
= ∑ ⎜ ⎟
⎝θ⎠ r
= ∑ ⎜⎜ r ⎟⎟
|r ⎝ θ ⎠
r =1 r =1
tr
⇒ Kr = coefficient of in KX(t)
|r
| r −1
= , r = 1, 2, 3,.......
θr
1 1
∴ K1 = , K2 =
θ θ2
2 6
K3 = , K4 =
θ3 θ4
Example: - Let X have the probability density function f(x) = θe−θx for x > 0 and zero,
otherwise. Prove that if the +ve part of x-axis is divided into intervals of equal length h starting
at the origin, then the probabilities that X will lie in successive intervals from a G.P. with
common ratio e−θh.
162
h h
Solution: - P(0 ≤ X ≤ h) = ∫ f ( x ) dx = ∫ θ e −θx dx
0 0
h
⎡ e −θx ⎤
=θ ⎢ −θh
⎥ = −1 e − 1 [ ]
⎣ − θ ⎦0
= 1− eθh
2h
∫θe
−θh
P(h ≤ X ≤ 2h) = dx
h
=−
θ
[ ]
θ −θx
e
2h
h [
= − e − 2 θh − e −θh ]
= e−θh ( 1−e−θh)
∞
⎡ e −2 x ⎤
= 2⎢ −4
⎥ = −1 0 − e = 4
1
[ ]
⎣ − 2 ⎦2 e
∞
(ii) Mean = μ1′ = E (X) = ∫ xf ( x ) dx
0
163
∞
= 2 ∫ x e −2 x dx
0
⎡⎛ e −2 x ⎞
∞
e −2 x ⎤
= 2 ⎢⎜⎜ x ⎟ + ∫ (1) dx ⎥
⎢⎝ − 2 ⎟ 2 ⎥
⎣ ⎠0 ⎦
⎡ 1
= 2 ⎢0 − 0 − e − 2 x
4
( ) ∞⎤
0 ⎥
⎣ ⎦
⎡ 1 ⎤ 1 1
= 2 ⎢− (0 − 1)⎥ = 2. =
⎣ 4 ⎦ 4 2
∞ ∞
μ ′2 = E (X 2 ) = ∫ x 2 f ( x ) dx = 2 ∫ x 2 e −2 x dx
0 0
⎡⎛ −2 x ⎞ ∞ ∞ −2 x ⎤ ⎡ ∞ ⎤
⎢ e e 1 ⎡1 1⎤ 1
μ ′2 = 2 ⎜ x 2 ⎟ + ∫ 2 x. dx ⎥ = 2 ⎢ 0 − 0 + .2 ∫ xe
−2x
dx ⎥ = 2 ⎢ . ⎥ =
⎢⎜ −2 ⎠ ⎟ +2 ⎥ ⎢⎣ 2 ⎥⎦ ⎣2 2⎦ 2
⎢⎣⎝ 0 0
⎥⎦ 0
1 1 1
Variance, μ2 = μ ′2 − μ1′ = − =
2 4 4
1
∴ S.D. = μ2 =
2
Mean 1 / 2
∴ = =1
S.D 1/ 2
∞
(iii) M.G.F., MX(t) = E[e ] = ∫ e tx f ( x ) dx
tx
∞ ∞ ∞
⎡ e ( t −2) x ⎤
= ∫ e (2)e
tx −2 x
dx = 2∫ e ( t −2 ) x
dx = 2⎢ ⎥
0 0 ⎣ ( t − 2) ⎦ 0
=
2
( t − 2)
[
e −( 2 − t ) x ]∞
0 =
2
t−2
[0 − 1] = − 2 = 2 , 2 > t
t−2 2−t
164
NORMAL DISTRIBUTION
A r.v. X is said to have Normal distribution with parameters μ and σ 2, called Mean & variance
respectively if its density function is given by
2
−1⎛ x −μ ⎞
1 ⎜ ⎟
f(x) = e 2⎝ σ ⎠
, − ∞ < x < ∞; σ > 0;−∞ < μ < ∞
σ 2π
X−μ x −μ
If Z= ⇒z=
σ σ
Here Mean = E(z) = 0
and variance = V(Z) = 1 = S.D.
1 2
1 −2z
then f(z) = e , −∞ < z < ∞
2π
then f(z) is known as Standard Normal distribution.
2
∞ ∞ − 1 ⎛⎜ x −μ ⎞⎟
1
∫ f (x) dx = σ 2π ∫e
2⎝ σ ⎠
Now dx
−∞ −∞
x −μ
Put z = ⇒ σ dz = dx
σ
∞ ∞ 1
1 − z2
∴ ∫ f (x ) dx = σ 2π ∫ e 2 .σ dz
−∞ −∞
∞ 1
1 − z2
=
2π
∫ e 2 dz
−∞
∞ ∞ 1
2 − z2
∴ ∫ f (x) dx = ∫ 2 dz
2π 0
e [ΘIntegrand is even function of z]
−∞
1 2
Put z =t
2
dt dr
⇒ z dz = dt ⇒ dz = =
z 2t
∞ ∞ ∞
2 dt 1
∫ f (x ) dx = ∫e ∫ e e dt
−t − t −1 / 2
∴ =
−∞
π 0 2π π0
165
1 ⎛1⎞ π
= . Γ⎜ ⎟ = =1 …(1)
π ⎝2⎠ π
⇒ it is a probability density function.
y
f(x)
−∞ x = 2μ x = −μ x=μ x ∞
∞ 1 2
− ( x −μ )
1 2
= ∫ x. e 2σ dx
−∞ σ 2π
x −μ
Put z= …(2)
σ
⇒ x = μ + σz ⇒ dx = σ dz
∞ 1
1 − z 2 σ dz
∴ μ1′ = ∫ (μ + σz). e 2
−∞ σ 2π
∞ 1 ∞ 1
μ − z2 σ − z2
=
2π
∫e 2 dz +
2π
∫ ze 2 dz
−∞ −∞
1
− z2
⇒ μ1′ = μ(1) + 0 [from(1)] [Θ integrand ze 2 is an odd function of z]
⇒ μ1′ = μ (Mean)
Variance, μ2 = E[(X−μ)2]
166
2
∞ 1 ⎛ x −μ ⎞
1 − ⎜ ⎟
= ∫ (x − μ) 2 e 2⎝ σ ⎠
dx
−∞ σ 2π
1
∞ − z2
1 e 2
∫σ
2
= z2 . σ dz
2π −∞
σ
∞ 1
2σ 2 − z2
∫z
2
= e 2 dz
2π 0
z2 dt
Put = t ⇒ dz =
2 2t
∞
2 2 dt
μ2 = σ ∫ 2t e −t
π 0 2t
∞
2 2 −t 1/ 2
= σ ∫ e t dt
π 0
2 2 ⎛3⎞ 2 2 1 ⎛1⎞
= σ Γ⎜ ⎟ = σ . Γ⎜ ⎟
π ⎝2⎠ π 2 ⎝2⎠
σ2
⇒ μ2 = . π = σ2 ⇒ μ2 = σ2
π
where x = 0, 1, 2,….., n
167
np
when x = 0 ⇒ z = − ; As n → ∞, z → −∞
q
n − np nq nq
when x = n ⇒ z = = =
npq npq p
and As n → ∞, z→∞
Thus in the limiting case as n→∞, z takes values −∞ to ∞.
Using Stirling’s approximation formula,
1
r+
−r
| r = 2π e r 2
1
n+
1 n 2 p x q n −x
= lim 1 1
2π n →∞ x+ n−x +
x 2 (n − x ) 2
1
n+
1 n 2 p x q n −x npq
= lim 1 1
.
2 π n →∞ x+ n −x + npq
x 2 (n − x ) 2
1 1
x+ n −x +
n +1
1 n p 2 q 2
= lim 1 1
2 π n →∞ x+ n −x +
npq x 2 (n − x ) 2
1 1
n −x + x+
n +1
1 n 1 (n.q) 2 (np) 2
= lim . 1 1 1
. 1
2π n →∞ npq x+ n −x + n −x + x+
x 2 (n − x) 2 n 2 n 2
1 1
x+ n −x +
1 1 ⎛ np ⎞ 2⎛ nq ⎞ 2
= lim ⎜ ⎟ ⎜ ⎟ …(3)
2 π n →∞ npq ⎝ x ⎠ ⎝n−x⎠
168
n npq q
∴ = 1+ z = 1+ z …(4)
np np np
n−x p
∴ =1−z …(5)
nq nq
⎛ 1⎞ ⎛ q ⎞ ⎛ ⎛ p ⎞
⎟ + ⎜ nq − z npq + ⎞⎟ log⎜1 − z
1
∴ log N = ⎜ np + z npq + ⎟ log⎜⎜1 + z ⎟ ⎜
⎟
⎝ 2⎠ ⎝ np ⎠ ⎝ 2⎠ ⎝ nq ⎟⎠
⎛ 1⎞ ⎡ q 1 2 q 1 3 q3/ 2 ⎤
= ⎜ np + z npq + ⎟ ⎢z − z + z .......⎥
⎝ 2 ⎠ ⎣ np 2 np 3 (np) 3 / 2 ⎦
⎛ 1⎞ ⎡ p z 2 p z3 p3/ 2 ⎤
+ ⎜ nq − z npq + ⎟ ⎢− z − − 3/ 2
.......⎥
⎝ 2⎠ ⎣ nq 2 nq 3 (nq ) ⎦
z2 z3 q3/ 2 z3 q3/ 2 z q z 2 q
= z npq − q+ + ...... + z 2 q − + −
z 3 nq 2 np 2 np 4 np
z 2 p z3 p3/ 2
− − + ........
4 nq 6 (nq ) 3 / 2
169
z2 1 ⎡ z 3 ⎛⎜ − q 3 / 2 p 3 / 2 ⎞⎟ z ⎛ q p ⎞⎤
logN = (p + q) + ⎢ ⎜ + + ⎜⎜ − ⎟⎥ + ......
2 n ⎢⎣ | 3 ⎝ p q ⎟⎠ 2 ⎝ p q ⎟⎠⎥⎦
z2 1 ⎡⎛⎜ z 3 p 2 − q 2 ⎞⎟ 3 ⎛⎜ q − p ⎞⎟⎤
= + ⎢ + ⎥ + [terms of higher powers 1/ n ]
2 n ⎢⎣⎜⎝ | 3 pq ⎟⎠ 2 ⎜⎝ pq ⎟⎠⎥⎦
Taking limit as n → ∞
lim log N= z 2 / 2
n →∞
1 2/ 2
or lim log N= z 2 / 2 ⇒ lim = e −z
n →∞ n →∞ N
Put this value in (6), we get the probability of the distribution of Z is given by
1 dx
dP = lim f ( x ) dx = lim .e −z / 2 npq dz [Θ dz = ]
n →∞ n →∞ 2π npq npq
1 −z 2 / 2
= e dz
2π
Hence probability function of Z is
1 −z2 / 2
f(z) = e , −∞ < z < ∞
2π
The probability function of normal distribution with variable X is
2
1 ⎛ x −μ ⎞
1 − ⎜ ⎟
f(x) = e 2⎝ σ ⎠
σ 2π
Some characteristics of Normal distribution
1 ⎛ x −μ ⎞ 2
1 − ⎜ ⎟ 0.5 0.5
f(x) = e 2⎝ σ ⎠
σ 2π
where −∞ < x < ∞
1
1 − 2 z2 x −μ −∞ x=μ−σ x=μ x=μ+σ +∞
and f(z) = e where z =
2π σ
(1) The curve is Bell shaped & symmetrical about the line x = μ
(2) Mean, Median & Mode coincides.
(3) Max. probability occurs at the point x = μ and is given by
170
1
[f ( x ) |max . =
σ 2π
(4) β1 = 0, β2 = 3
(5) All odd order moments are zero i.e. μ2n+1 = 0
(6) Linear combination of independent normal variate is also a normal variate.
(7) Points of inflexion of the curve are given by x = μ ± σ
4
(8) Mean Derivation is σ (appox.)
5
2
= σ
π
(9) Area property
P(μ −σ < × < μ + σ) = 0.6826
P(μ−2σ < × < μ + 2σ) = 0.9544
P(μ−3σ < × μ + 3σ) = 0.9973
0.6826
171
f ′( x ) = 0 , f ′′( x ) < 0
2
1 ⎛ x −μ ⎞
1 − ⎜ ⎟
2⎝ σ ⎠
Now f(x) = e …(1)
σ 2π
2
1⎛ x −μ ⎞
∴ log f(x) = − log σ 2π − ⎜ ⎟
2⎝ σ ⎠
Differentiating w.r.t. x, we get
1 2 −1
f ′( x ) = 0 − 2 ( x − μ ) = 2 ( x − μ )
f (x) 2σ σ
f (x)
⇒ f ′( x ) = − (x − μ) …(2)
σ2
Now f ′( x ) = 0 ⇒ x − μ = 0 ⇒ x = μ
−1
Also f ′′( x ) = [1.f (x ) + (x − μ)f ′(x )]
σ2
−1 ⎡ f ( x )( x − μ ) ⎤
= 2 ⎢
f (x) − (x − μ) ⎥⎦ from(2)
σ ⎣ σ2
f (x) ⎡ (x − μ) 2 ⎤
= ⎢1 − ⎥
σ2 ⎣ σ2 ⎦
Put x = μ, we get
f ′′( x ) = −
1
2
[(f ( x )]x =μ = − 21 . 1 < 0 [By using (1)]
σ σ σ 2π
Hence x = μ is the mode of Normal distribution.
Median
Median is the middle value of variate. Let M be the Median distribution, then
∞ ∞
∫ f (x) dx + ∫ f (x) dx = 1
−∞ M
M M
Also ∫ f (x) dx = ∫ f (x) dx
−∞ −∞
M
1
So we have ∫ f (x ) dx = 2
−∞
172
2
M 1 ⎛ x −μ ⎞
1 − 2 ⎜⎝ ⎟ 1
∫
σ ⎠
⇒ e dx =
−∞ σ 2 π
2
1 ⎛ x −μ ⎞ 2 1 ⎛ x −μ ⎞2
μ − ⎜ ⎟ − ⎜M ⎟
1 1 1
∫e ∫ ⎝
2⎝ σ ⎠ 2 σ ⎠
⇒ dx + e dx =
σ 2π −∞ σ 2π μ 2
1
or I1 + I 2 =
2
1 ⎛ x −μ ⎞
1 − ⎜ ⎟
∫e
2⎝ σ ⎠
Now I1 =
σ 2π
x −μ
Put z=
σ
0
1 −z 2 / 2
∴ I1 =
σ 2π
∫e dz
−∞
0
1 −z 2 / 2
=
2π
∫e dz
−∞
∞ 1
1 − z2
⇒ I1 =
2π
∫e 2 dz
0
dt
Put z2/2 = t ⇒ z dz = dt ⇒ dz =
2t
∞
1 1 −1/ 2
I1 = ∫
2π 0
e −t
2
t dt
1 ⎛1⎞ 1 1
= Γ⎜ ⎟ = . π=
2π ⎝2⎠ 2 π 2
μ 1 ⎛ x −μ ⎞2
1 − ⎜ ⎟
∫e
2⎝ σ ⎠
⇒ I2 = 0 ⇒ dx = 0
σ 2π μ
⇒ M=μ
Hence Mean = Madian = Mode
173
Moment generating function about origin
∞
MX(t) = E[etx] = ∫e
tx
f ( x ) dx
−∞
2
∞ 1 ⎛ x −μ ⎞
1 − ⎜ ⎟
∫e
tx 2⎝ δ ⎠
= e dx
σ 2π −∞
x −μ
Put z=
σ
⇒ x = μ + σz
dx = σ dz
∞ 1
1 −
∫e
t ( μ + σz )
MX(t) = e 2 .σ dz
σ 2π −∞
∞ 1
e μt − ( z 2 − 2 tσz )
Mx(t) =
2π
∫ e 2 dz
−∞
2 2 1 2 2
μt ∞ 1 2 + t σ − 2 tσz ) + 2 t σ
e − (z
=
2π
∫ e 2 dz
−∞
∞ 1 1 2 2
e μt − ( z − tσ ) 2 t σ
=
2π
∫e 2 e2 dz
−∞
1 2
μt + t 2 σ ∞ 1
e 2 − ( z − tσ ) 2
=
2π
e 2 ∫ dz
−∞
Put z−tσ = μ ⇒ dz = dμ
1
μt + t 2σ 2 ∞ 1
e 2 − μ2
∴ MX(t) =
2π
∫
e 2 dμ
−∞
1
μt + t 2σ 2 ⎡ 1
∞ 1
− z2 ⎤
= e 2 (1) ⎢Θ
⎢⎣ 2π
∫ e 2 dz = 1⎥
⎥⎦
−∞
1
μt + t 2σ 2
=e 2 …(1)
174
Moment generating function about mean
MX−μ(t) = E[et(X−μ)] = E[etX.e−μt]
= e−μt E[etX] = e−μt MX(t)
1 22 1 22
μt μt + σ t σ t
= e− e 2 = e2
Also by defination
2
∞ 1 ⎛ x −μ ⎞
1 − ⎜ ⎟
MX−μ(t) = ∫ e t ( x −μ ) e 2 ⎝ σ ⎠
dx
−∞ σ 2 π
2
∞ 1 ⎛ x −μ ⎞
1 − ⎜ ⎟
= e ∫ σ 2π
tx 2⎝ σ ⎠
−μt e e dx
−∞
1 1 22
μt + σ 2 t 2 σ t
−μt
=e .e 2 = e2 [from(1)]
Moment generating function about any point ‘a’
1
t ( μ −a ) + σ 2 t 2
MX−a(t) = e 2
By defination
MX−a(t) = E[et(X−a)]
2
∞ 1 ⎛ x −μ ⎞
1 − ⎜ ⎟
∫e
t ( x −a ) 2⎝ σ ⎠
MX−a(t) = e dx
σ 2π −∞
2
∞ 1 ⎛ x −μ ⎞
1 ⎜ ⎟
= e −at . ∫e
tx 2⎝ σ ⎠
.e dx
σ 2π −∞
1 1
μt + σ 2 t 2 t ( μ −a ) + σ 2 t 2
= e −at . e 2 =e 2
Cumulants
⎛ μt + 1 σ2t 2 ⎞
KX(t) = log MX(t) = log ⎜ e 2 ⎟
⎜ ⎟
⎝ ⎠
1 2 2
= μt + σ t
2
The 1st cumulants, K1 = coefficient of t = μ1′ (Mean) = μ
175
The 2nd cumulants, K2 = σ2 = variance = μ2
Kr = 0 for r = 3, 4,…….
Thus μ3 = K3 = 0 and μ4 = K4 + 3 K 22 = 3σ4 [Θ K4 = μ4− K 22 ]
μ 4 3σ 2
Now β1 = = 4 =3 (β1 = 0, β2 = 3)
μ2 σ
tr
(1), μr = coefficient of = rth moment about mean
|r
Since there is no term with odd powers of t in (1), there fore all moments of odd order about
Mean vanish.
i.e. μ2n+1 = 0, n = 1, 2, ……
t 2n σ 2 n | 2n
and μ2n = coefficient of in (1) =
| 2n 2n | n
⎡ t 2 n σ 2 n .| 2n ⎛ σ 2 n | 2n ⎞ t 2n ⎤
⎢Θ n =⎜ n ⎟ ⎥
⎢⎣ 2 | n | 2n ⎜ 2 |n ⎟ | 2n ⎥
⎝ ⎠ ⎦
σ 2n
∴ μ2n = [2n (2n − 1) (2n − n ).....4.3.2.1]
2n | n
σ 2n
= [1.3.5…….(2n−1)] [2.4.6.(2n−2)2n]
2n | n
σ 2n
= [1.3.5…….(2n−1) 2n [1.2…..n]
2n | n
σ 2n
= n
[1.3.5…….(2n−1) 2n | n
2 |n
176
⇒ μ2n = σ2n[ 1.3.5……(2n−1)]
∴ μ2 = 1.σ2 = σ2
μ3 = 0, μ4 = 1. 3σ4
μ5 = 0, μ6 = 1.3.5σ6
β1= 0, β2 = 3
Also μ1′ = moment about origin
∞
∫ x f (x) dx
r r
= E[X ] =
−∞
∫ (x − μ)
2 n +1
= f ( x ) dx
−∞
∫ (x − μ)
2n
and μ2n = f ( x ) dx = even order moments about mean
−∞
Characteristic Function
1
iμt − t 2σ 2
itx
φX(t) = E[e ] = e 2
Mean deviation
∞
M.D = ∫ | x − μ | f (x ) dx
−∞
2
∞ 1 ⎛ x −μ ⎞
1 − ⎜ ⎟
= ∫ | x −μ | e 2⎝ σ ⎠
dx
−∞ σ 2π
x −μ 1
Put z= ⇒ dz = dx
σ σ
1
∞ − z2
1 e 2
=
2π
∫ | σz | σ
.σ dz
−∞
∞ 1
2 − z2
∴ M.D =
2π
∫ σz e 2 dz [Θ integrand is eveb function of z]
0
177
1 2 1
Put z =t ⇒ .2z dz = dt
2 2
⇒ z dz = dt
∞
2
∴ M.D. = σ ∫ e − t dt
2π 0
∞
2 ⎡ e −t ⎤ 2 2
= σ ⎢ ⎥ = σ.[0 − 1] = σ
π ⎣ −1 ⎦0 π π
4 2
= σ = 0.8 σ ⇒ M.D = σ
5 π
f (x) ⎡ (x − μ) 2 ⎤
Now f ′′( x ) = − ⎢1 − ⎥
σ2 ⎣ σ2 ⎦
(x − μ) 2
Put f ′′( x ) = 0, 1− =0
σ2
⇒ (x−μ)2 = σ2 ⇒ x−μ = ±σ
⇒ x=μ±σ
f ′( x ) ⎡ ( x − μ ) 2 ⎤ f ( x ) ⎡ − 2( x − μ ) ⎤
Now f ′′′( x ) = − ⎢1 − ⎥−
σ2 ⎣ σ 2 ⎦ σ 2 ⎢⎣ σ2 ⎥⎦
f ′( x ) ⎡ ( x − μ ) 2 ⎤ 2f ( x )( x − μ )
=− ⎢1 − ⎥+
σ2 ⎣ σ2 ⎦ σ4
At x = μ ± σ,
⎡ (μ + σ − μ ) 2 ⎤ 2f ( x )
− f ′( x )
f ′′′( x ) = ⎢1 − ⎥ + 4 (μ + σ − μ )
σ2 ⎣ σ2 ⎦ σ
− f ′( x )
⇒ f ′′′( x ) = 2
[1 − 1] + 2f (3x ) .σ
σ σ
2f ( x ) 2f (μ + σ)
=0+ = ≠0
σ3 σ3
At x = μ−σ,
178
f ′( x ) ⎡ (−σ) 2 ⎤ 2f ( x )
f ′′′( x ) = − ⎢1 − 2 ⎥ + 4 (−σ)
σ2 ⎣ σ ⎦ σ
f ′( x ) 2f ( x ) 2f ( x )
= − 2 (1 − 1) − 3 = 0 − 3
σ σ σ
− 2f ( x ) − 2f (μ − σ)
⇒ f ′′′( x ) = = ≠0
σ3 σ3
(±σ) ⎡ (±σ) 2 ⎤
′′′
f ( x ) = 4 f (μ ± σ) ⎢3 − 2 ⎥
σ ⎢⎣ σ ⎥⎦
± 2f (μ ± σ)
= ≠0
σ3
The S.D. is the distance of point of inflexion from the axis of symmetry.
μ ′4 (about 50) = 48
Find Mean and S.D.
Solution: - By defination 1st moment about 10,
∞
μ1′ = ∫ ( x − 10)f ( x ) dx = 40
−∞
∞ ∞
⇒ = ∫ xf (x ) dx − 10 ∫ f (x) dx = 40
−∞ −∞
∞ ⎡ ∞ ⎤
⇒ = ∫ xf (x ) dx = 40 + 10 = 50 ⎢Θ ∫ f ( x ) dx = 1⎥
−∞ ⎢⎣ −∞ ⎥⎦
∞
∴ Mean = ∫ xf (x ) dx = 50
−∞
As Mean is 50.
179
μ ′4 = 48(4th moment about Mean)
⇒ 3σ4 = 48 ⇒ σ4 = 16 = 24
⇒ σ = S.D. = 2
Theorem: - If X1 and X2 are independent normal variates, then X1 + X2 is also a normal variate.
Proof: - X1 ∼ N(μ1, σ12 )
X2 ∼ N(μ2, σ 22 )
[ ]= e
1
μ1t + σ12 t 2
M X1 ( t ) = E e tX1 2
[ ]
1
μ 2t + σ2 t2
M X 2 ( t ) = E e tX 2 = e 2 2
which is the M.G.F of Normal distance with Mean μ1+μ2 and variance σ12 + σ 22
1 1 2 22
μ1a1t + σ12a12 t 2 μ 2a 2 t + 2 σ 2a 2 t
= e 2 .e .......... [from(1)]
180
t (∑ a iμ i ) + t 2 ⎛⎜ ∑ a i2 σi2 ⎞⎟
1
=e 2 ⎝ ⎠
1 ⎧1 n
1 n ⎫
then (X1 + X2 + ……+ Xn) ∼ N ⎨ ∑μi , n2
∑ σ i2 ⎬
n ⎩n i =1 i =1 ⎭
⎧1 1 ⎫
⇒ Xi ∼ N ⎨ .nμ, 2 nσ 2 ⎬
⎩n n ⎭
[ΘAs each Xi has mean = μ, variance = σ2]
⎛ 1 ⎞
⇒ Xi ∼ N ⎜ μ, σ 2 ⎟
⎝ n ⎠
X − np 2 /2
Theorem: - If X ∼ B (n, p), then prove that MGF of tends to e t as n→∞.
npq
X − np
Proof: - Let U =
npq
∴ log MU(t) =
− npt
npq
+ n log q + pe t / [ npq
]
− npt ⎡ ⎧⎪ t t2 t3 ⎫⎪⎤
= + n log ⎢1 + p⎨ + + 3/ 2
+ ....... ⎬⎥
npq ⎢⎣ ⎪⎩ npq | 2 npq | 3 (npq) ⎪⎭⎥⎦
t2
= (1 − p) + 0(n −1 / 2 )
2q
t2 ⎛ 1 ⎞
= + 0⎜ ⎟
2 ⎝ n⎠
As n→∞, log MU(t) = t2/2
2 /2
⇒ MU(t) = e t
Area property
If X ∼ N(μ, σ2)
The probability that r.v. X will lie between X = μ and X = x1 is given by
x1
P(μ < X < x1) = ∫ f (x) dx
μ
x 1 ⎛ x −μ ⎞ 2
1 1 − 2 ⎜⎝ ⎟
∫e
σ ⎠
= dx
σ 2π μ
X−μ x −μ
Put Z= , then z =
σ σ
If X = μ, then Z = 0
x1 − μ
& when X = x1, Z = z1(say) where z1 =
σ
∴ P(μ < X < x1) = P( 0 < Z < z1)
182
z1 1 2 z1
1 − z
=
2π
∫ e 2 dz = ∫ φ(z) dz
0 0
1 −z2 / 2
where φ(z) = e is the probability function.
2π
z1
The definite integral ∫ φ( z ) dz is known as normal probability Integral & gives the area under
0
Normal curve between ordinates at Z = 0& Z = z1. These are tabulated for different values of z1
at interval of 0.01
= 2(0.3413)
= 0.6826
and P(μ − 2σ < X < μ + 2σ) = P(−2 < Z < 2)
2 2
= ∫ φ(z) dz = 2∫ φ(z) dz
−2 0
= 2(0.4772) = 0.9544
and (μ − 3σ < X < μ + 3σ) = P(−3 < Z < 3)
= 2(0.4987) = 0.9974
and outside rangeμ ± 3σ
183
P(|X−μ| > 3σ) = P(|Z| > 3)
= 1−P(−3 < Z < 3)
= 1−0.9974 = 0.0026
Example: - If X ∼ N(12, 16)
Find (i) P(X ≥ 20) (ii) P(X ≤ 20) (iii) P(0 ≤ X ≤ 12)
Solution: - (i) Here μ = 12, σ = 4
X − μ X − 12
Z= =
σ 4
∴ P(X ≥ 20) = P(Z ≥ 2)
= 0.5 − P(0 ≤ Z < 2)
= 0.5 − 0.4772 = 0.0228
(ii) P(X ≤ 20) = 1− P(X ≥ 20)
= 1−0.0228 = 0.9772
(iii) P(0 ≤ X ≤ 12) = P(−3 ≤ Z ≤ 0) = P(0 ≤ Z ≤ 3) (due to symmmtrical)
∴ P( 0 ≤ X ≤ 12) = 0.4987
(iv) Find x` when P(X > x`) = 0.24
x ′ − 12
Solution: - when X = x`, then Z = = z1 (say) …(1)
4
∴ P(Z > z1) = 0.24
P(0 < Z < z1) = 0.5 − 0.24 = 0.26
From table, z1 = 0.71
x′ − 12
∴(1)⇒ z1 = = 0.71
4
⇒ x ′ = 14.84
μ σ2
Example: - X ∼ N( 30, 25 )
184
40 − 30
& when X = 40, Z= = 2.0
5
∴ P(26 ≤ X ≤ 40) = P( − 0.8 ≤ Z ≤ 2)
= P(− 0.8 ≤ Z < 0.8) + P(0 ≤ Z ≤ 2)
= P(0 ≤ Z ≤ 0.8) + P(0 ≤ Z < 2)
= 0.2881 + 0.4772
P(26 ≤ X ≤ 40) = 0.7653
Find x10 and x 11 when P( x10 ≤ X ≤ x 11 ) = 0.50 and P(X > x 11 ) = 0.25
Solution: -
0.5
0.25
0.25
0.25
x=x0′ x=x1′
z=z1 z=z1
z=0 185
x 1′ − 12
when X = x 1′ , Z= = z1 (say)
4
x ′0 − 12
when X = x ′0 , Z= = −z1 (from figure)
4
P(X > X1′ ) = 0.25 ⇒ P( Z > z1) = 0.25
x ′0 − 12
and = −0.67 ⇒ x ′0 = 9.32
4
186
LESSON 8 MULTIPLE AND PARTIAL CORRELATION
INTRODUCTION.
When the value of the one variable is associated with or influenced by other
variable, e.g., the age of husband and wife, the height of father and son, the supply
and demand of a commodity and so on, Karl Pearson’s coefficient of correlation
can be used as a measure of linear relationship between them. But sometimes there
is correlation between many variables and the value of one variable may be
influenced by many others, e.g., the yield of crop per acre say (X1) depends upon
quality of seed (X2), fertility of soil (X3), fertilizer used (X4), irrigation facilities
(X5) etc. Whenever we are interested in studying the joint effect of a group of
variable upon a variable not included in that group , our study is that of multiple
correlation and multiple regression. The correlation and regression between only
two variates after eliminating the linear effect of other variates in them is called the
partial correlation and partial regression.
Yule’s Notation:
Without loss of generality, we can assume the variables X1, X2, and X3 have been
measured from their respective means, so that
187
The coefficient b12.3 and b13.2 are known as the partial regression coefficients of
X1 on X2 and of X1 on X3 , respectively. The quantity
is called the estimate of X1 as given by the plane of regression (1) and the quantity.
In the general case of n variable X1, X2,…… Xn the equation of the plane of
regression of X1 on X2, X3,……. Xn becomes
The notation used here are due to Yule. The subscripts before the dot (.) are
known as primary subscripts and those after the dot are called secondary
subscripts. The order of regression coefficient is determined by the number of
secondary subscripts, e.g.,
188
The order of the residual is also determined by the number of secondary
subscripts in it, e.g., X1.23, X1.234,…….X1.23…..n, are the residual of order 2,3,…….
(n-1) respectively.
Remaks: In the following sequences we shall assume that the variable under
consideration has been measured from their respective means.
Plane of regression:
The constants b’s in (3) are determined by the principle of leastsquares, i.e., by
minimizing the sum of the squares of the residual, viz.,
the summation being extended to the given values (N in number) of the variables.
∂S ⎫
= 0 = −2∑ X 2 (X 1 − b12.3 X 2 − b13.2 X 3 )⎪
∂b12.3 ⎪
⎬ (4)
∂S
= 0 = −2∑ X 3 (X 1 − b12.3 X 2 − b13.2 X 3 ) ⎪
∂b13.2 ⎪⎭
⇒ ∑X X 1 2 − b12.3 ∑ X 22 − b13.2 ∑ X 2 X 3 ⎫⎪
2⎬
(6)
∑X X 1 3 − b 12.3 ∑ X 2 X 3 − b 13.2 ∑ X 3 ⎪
⎭
189
1 1 ⎫
σ i2 =
N
∑ X i2 , Cov(X i , X j ) = ∑ X i X j ⎪
N ⎪
Cov(X i , X j ) ∑ X i X j ⎬ (7)
rij = = ⎪
σiσ j Nσ i σ j ⎪⎭
Similarly, we obtain
1 r12
σ1 r23 r13
b13.2 = (10)
σ3 1 r23
r23 1
If we write
1 r12 r13
ω = r21 1 r23 (11)
r31 r32 1
And ωij is the cofactor of the element in the ith row and jth column of ω, we have
from (9) and (10)
σ1 ω12 σ ω
b12.3 = − and b13.2 = − 1 13 (12)
σ 2 ω11 σ 3 ω11
190
Substituting thes values in (3), we get the required equation of the plane of
regression of X1 on X2 and X3
σ1 ω12 σ ω
X1 = − . .X 2 − 1 13 .X 3 = 0
σ 2 ω11 σ 3 ω11
(13)
X1 X X
⇒ .ω11 + 2 .ω12 + 3 .ω13 = 0
σ1 σ2 σ3
or Eliminating the coefficient b12.3 and b13.2 in (3) and (8), the required equation of
the plane of regression of X1 on X2 and X3 becomes
X1 X2 X3
r12 σ1σ 2 σ 2
2 r23 σ 2 σ 3 = 0
r13 σ1σ 3 r23 σ 2 σ 3 σ 32
Dividing C1, C2 and C3 by σ1, σ2 and σ3 respectively and also R2 and R3 by σ2 and σ3
respectively, we get
X1 X2 X3
σ1 σ2 σ3
r12 1 r23 = 0
r13 r23 1
X1 X X
⇒ .ω11 + 2 .ω12 + 3 .ω13 = 0
σ1 σ2 σ3
Generalization:
191
S = ∑ X 12.23......n
Using the principle of least squares, the normal equations for estimating the
(n-1), b’s are
∂S ⎫
= 0 = −2∑ X 2 (X 1 − b12.34..........n X 2 − b13.24..........n X 3 − ..........b1n .23...........( n −1) X n ) ⎪
∂b12.34.......n
⎪
∂S ⎪
= 0 = −2∑ X 3 (X1 − b12.34.........n X 2 − b13.24..........n X 3 − ...........b1n .23...........( n −1) X n ) ⎪
∂b13.24.........n
⎪
. ⎪
⎪
. ⎬
. ⎪
⎪
. ⎪
∂S ⎪
= 0 = −2∑ X n (X 1 − b12.34.........n X 2 − b13.24..........n X 3 − ...........b1n .23...........( n −1) X n )⎪
∂b1n .23.........n ⎪
⎪
⎭
(15)
. . .
. . .
. . .
192
Hence, the elimination of b’s between (14) and (15) yields
X1 X2 X3 ............ Xn
r12 σ1σ 2 σ 22 r23 σ 2 σ 3 ............. r23 σ 2 σ 3
r13 σ1σ 3 r23 σ 2 σ 3 σ 32 .............. r3n σ 3 σ n = 0
Μ Μ Μ Μ Μ
r1n σ1σ n r2 n σ 2 σ n r3n σ 3 σ n Κ ......... σ 2n
Dividing C1, C2 …….. ,Cn by σ1, σ2…….σn respectively and also R2, R3…..Rn by
σ2, σ3…. σn respectively, we get
X1 X2 X3 Xn
........
σ1 σ2 σ3 σn
r12 1 r32 ......... r2 n
=0 (18)
r13 r23 1 ........ r3n
Μ Μ Μ ........ Μ
r1n r2 n r3n ......... 1
If we write
And ωij is the cofactor of the element in the ith row and jth column of ω, we get
from (18)
X1 X X X
⇒ .ω11 + 2 .ω12 + 3 .ω13 + ................ + . n .ω1n = 0 (20)
σ1 σ2 σ3 σn
193
σ1 ω12 σ ω σ ω
X1 = − . .X 2 − 1 13 .X 3 − .......... 1 1n .X n = 0 (21)
σ 2 ω11 σ 3 ω11 σ n ω11
σ1 ω12 ⎫
b12.34 =−
σ 2 ω11 ⎪
⎪
σ1 ω13 ⎪
b13.24 =− ⎪
σ 3 ω11 ⎬ (22)
Μ Μ ⎪
σ ω ⎪
b1n .23......( n −1) = − 1 1n ⎪
σ n ω11 ⎭⎪
Properties of residuals
Property 1. The sum of the product of any residual of order zero with any other
residual of higher order is zero, provided the subscript of the former occurs among
the secondary subscripts of the latter.
∑ X2 X1.23 = 0, ∑ X3 X1.23 = 0
Property 2. The sum of the product of any two residual in which all the secondary
subscripts of the first occur among the secondary subscripts of the second is
unaltered if we omit any or all the secondary subscripts of the first.
194
Property 3. The sum of the product of two residuals is zero if all the subscripts
(primary as well as secondary) of the one occur among the secondary subscripts of
the other, e.g.,
∑ X1.2 X3.12 = 0
Since all the Xi’s are measured from their respective means, we have
1 1
σ12.23....n =
N
∑ [X 1.23....n − E (X 1.23....n )]2 = ∑ X12.23....n
N
1 1
σ12.23....n =
N
∑ X 1.23....n X 1.23....n = ∑ X 1 X1.23....n
N
1
=
N
∑ X1 (X1 − b12.34....n X 2 − b13.24....n X 3 − .........b1n.23....( n −1) X n 2
= σ12 − b12.34....n r12 σ1σ 2 − b13.24....n r13 σ1σ 3 − ............... − b1n .23........( n −1) r1n σ1σ n
⇒ σ12 − σ12.23......n = b12.34....n r12 σ1σ 2 − b13.24....n r13 σ1σ 3 − ............... − b1n .23........( n −1) r1n σ1σ n
Eliminating the b’s we get
195
σ12 − σ12.23......n r12 σ1σ ...... r1n σ1σ n
r12 σ1σ σ 22 ...... r2 n σ 2 σ n
= 0
Μ Μ Μ Μ
r1n σ1σ n r2 n σ 2 σ n ..... σ 22
Dividing R1, R2…..Rn by σ1, σ2…. σn respectively and also C1, C2 …….. ,Cn by σ1,
σ2…….σn respectively, we get
σ12.23......n
1− − r12 ...... r1n
σ12
r12 1 ...... r2 n = 0
Μ Μ Μ Μ
r1n r2 n ..... 1
σ12.23......n
1 r12 ...... r1n r12 ...... r1n
σ12
r12 1 ...... r2 n
− 0 1 ...... r2 n = 0
Μ Μ Μ Μ
Μ Μ Μ Μ
r1n r2 n ..... 1
0 r2 n ..... 1
σ12.23......n
⇒ ω− ω11 = 0
σ12
ω
∴ σ12.23......n = σ12 − (23)
ω11
ω
∴ σ12.23......n = σ12 − (24)
ω11
196
Coefficient of Multiple Correlation:
we have,
⇒ e1.23 = X1 - X1.23
By definition
Cov(X 1 , e1.23 )
R 1.23 = (25)
V(X 1 )V(e1.23 )
= σ12 –σ1.232
Also
σ12 − σ12.23
R 1.23 =
σ12 (σ12 − σ12.23 )
197
σ12.23
1− R 2
1.23 = 2 (26)
σ1
Using (24), we get
ω
1 − R 12.23 = (27)
ω11
1 r12 r13
Where, ω = r21 1 r23
r31 r32 1
1 r23
and ω11 =
r32 1
ω r 2 + r 2 − 2r r r
R 12.23 = 1 − = 12 13 2 12 13 23
ω11 1 − r23
This formula expresses the multiple correlation coefficient in terms of the total
correlation coefficients between the pairs of variables
Generalisation:
σ12 − σ12.23..........n
R 1.23........n =
σ12 (σ12 − σ12.23........n )
σ12.23........n
R 12.23.......n = 1 −
σ12
ω
R 12.23......n = 1 − (28)
ω11
198
Coefficient of Partial Correlation:
We have
1 1
Cov(X 1.3 , X 2.3 ) =
N
∑ X1.3 X 2.3 = ∑ X 1 X 2.3
N
1 1
=
N
∑ X 1 (X 2 − b 23 X 3 ) = ∑ X 1 X 2 − b 23 ∑ X 1 X 3
N
1 2 1
Var (X1.3 ) = ∑ X1.3 = ∑ X1.3 X 2.3
N N
1 1
= ∑ X1X 2.3 = ∑ X1 (X1 − b13 X 3 )
N N
1 1
= ∑ X12 − b13 . ∑ X1X 3
N N
σ
= σ12 − r13 1 r13σ1σ 3
σ3
= σ12 (1 − r13
2
)
199
Example: From the data relating to the yield of grain (X1),fertilizer used (X2) and
whether condition (X3) for 18 crops the following correlation coefficients were
obtained:
Fond the partial correlation coefficient r12.3 and multiple correlation coefficient R1.23
Solution:
R 12.23 = = = 0.7334
1 − r232 1 − (0.52) 2
R1.23 = +0.8564
E(Xi) = μ1
200
Proof: Let M1(t) denote the M.G.F. of each of the derivation (Xi-μ1) and M(t)
denote the M.G.F. of the standard variate
Z = (Sn-μ)/σ
′ ′
Since μ1 = E(X i − μ1 ) = 0, μ 2 = E(X i − μ1 ) 2 = σ12
We have
⎛ ′ ′t
2
′t
2
⎞
⎜
M 1 ( t ) = ⎜1 + μ 1 t + μ 2 + μ3 + ........⎟⎟
⎝ 2 6 ⎠
(1)
⎡ t 2
⎤
= ⎢1 + σ12 + O( t 3 )⎥
⎣ 2 ⎦
We have
S n − μ (X 1 + X 2 ....... + X n n
⎛ X − μ1 ⎞
Z= = = ∑⎜ i ⎟
σ σ i =1 ⎝ σ ⎠
M z (t) = M n ⎛ X − μ ⎞ (t) = M n (σ / t )
i 1 ∑ X i − μ1
∑⎜ ⎟
i =1⎝ σ ⎠ i =1
{ } [ ]
n
= ∏ M ( X i −μ1 ) ( t / σ) = [M 1 ( t / σ)] = M ( X i −μ1 ) ( t / n σ1
n n
i =1
n From (1)
⎡ t2 ⎤
= ⎢1 + + O( n − n / 2 ) ⎥
⎣ 2n ⎦
For every fixed ‘t’, the term O(n-3/2)→0 as n→∞. Therefore, as n→∞, we get
201
n 2
t
⎡ t2 −n / 2
⎤ ⎡t2 ⎤
lim it M z ( t ) = lim ⎢1 + + O( n )⎥ = exp ⎢ ⎥ = e 2 ,
n →∞ n →∞ ⎣⎢ 2n ⎦⎥ ⎣⎢ 2 ⎦⎥
Let X1 , X2,………. ,Xn be a sequence of random variables and μ1, μ2, ……, μn
⎧ X + X 2 + ...... + X n μ1 + μ 2 + ........μ n ⎫
Then P ⎨ 1 − ≤ ε⎬ ≥ 1 − η
⎩ n n ⎭
For all n > n0, where ε and η are arbitrary small positive numbers, provided
Bn
lim →0
n →∞ n 2
⎧ X + X 2 + ...... + X n X + X 2 + ...... + X n ⎫ B
P⎨ 1 − E( 1 ) ≤ ε⎬ ≥ 1 − n
⎩ n n ⎭ n 2ε 2
⎡ ⎛ X1 + X 2 + ...... + X n ⎞⎤ 1 B
⎢Since Var⎜ ⎟⎥ = 2 Var (X 1 + X 2 + ...... + X n ) = 2n
⎣ ⎝ n ⎠⎦ n n
202
⎧ X + X 2 + ...... + X n ⎛ μ1 + μ 2 + ...... + μ n ⎞ ⎫ B
⇒ P⎨ 1 −⎜ ⎟ < ε⎬ ≥ 1 − 2 n 2
⎩ n ⎝ n ⎠ ⎭ n ε
Bn
Since ε is arbitrary, we assume →∞, as n becomes indefinitely large. Thus,
n 2ε 2
having chosen two arbitrary small positive numbers ε and η, number n0 can be
found so that the inequality
Bn
< η, is 1
n 2ε 2
⎧ X + X 2 + ...... + X n μ1 + μ 2 + ........μ n ⎫
P⎨ 1 − ≤ ε⎬ ≥ 1 − η .
⎩ n n ⎭
Chi-Square Distribution
X−μ
Definition. If X ~ N(μ, σ2) , then Z = ~ N(0, 1)
σ
2
⎛X−μ⎞
And Z = ⎜
2
⎟ , is the chi-square variate with 1 d. f
⎝ σ ⎠
2
n
⎛ X − μi ⎞
χ = ∑ ⎜⎜ i
2
⎟⎟ , is the chi-square variate with n d. f
i =1 ⎝ σi ⎠
203
2
⎛ X − μi
n
⎞ n
χ = ∑ ⎜⎜ i
2
⎟⎟ = ∑ U i2
i =1 ⎝ σi ⎠ i =1
Xi − μi
Where U i =
σi
(1 )
dP(χ ) = 2
2
n/2
⎡ 1 ⎤
.⎢exp(− χ 2 )⎥ (χ 2 ) ( n / 2 ) −1 dχ 2
Γ(n / 2) ⎣ 2 ⎦ 0 ≤ χ 2 <∞
1 ⎡ χ 2
⎤
= .⎢exp(− )⎥ (χ 2 ) ( n / 2) −1 dχ 2 ,
2 n/2
Γ(n / 2) ⎣ 2 ⎦
Students t distribution
x −μ
t= n,
s
204
1
f (t ) = -∞<t<∞.
ν +1
1 ν t2 2
ν B( , ){1 + }
2 2 ν
Moments
Mode=mean =median=0
∞ 2 ∞ t 2r
μ2r = μ2r’ = ∫ t 2r f ( t )dt = ∫ dt
1 ν 0 ν +1
−∞ ν B( , ) t2 2
2 2 {1 + }
ν
ν 1
ν r B( − r, r + )
= 2 2
1 ν
B( , )
2 2
n
Put r=1, then μ2 = , n>2,
n−2
is known as variance.
F –distribution.
205
U
( ) U ~ χ ν2
ν1 1
F= , (ν1> ν2), | as
V V ~ χ ν2
( )
ν2 2
ν1
ν ν1
( 1) 2 −1
ν2 F2
f (F) = , 0<F<∞.
ν1 ν 2 ν1 + ν 2
B( , ) ν
2 2 {1 + 1 F} 2
ν2
Moments.
ν1
r ν1 2 ν1
−1
F ( )
∞ ∞ ν2 F2
‘ r r
μr =E(F ) = ∫ F f (F)dF = ∫ dF
0 B( ν1 , ν 2 )
ν1 + ν 2
0 ν1
2 2 {1 + F} 2
ν2
ν
( 2 )r
ν1 ν ν
μ 'r = B(r + 1 , 2 − r )
ν ν 2 2
B( 1 , 2 )
2 2
ν2
Putting r=1, we get mean = μ1' =
ν2 − 2
ν 2 2 (ν1 + 2)
Further μ '2 = , ν 2 >4
ν 2 (ν 2 − 2)(ν 2 − 4)
206
ν 2 2 ⎡ ν12 − 4 − ν1ν 2 + 4ν1 ⎤
Now variance is obtained from μ 2 = μ '2 − μ1'2 = ⎢ ⎥
(ν 2 − 2) ⎢⎣ ν1 (ν 2 − 4)(ν 2 − 2) ⎥⎦
207