Sunteți pe pagina 1din 22

Universit`a di Pavia

Introduction to Stochastic processes


Eduardo Rossi
Stochastic Process
Stochastic Process: A stochastic process is an ordered sequence of
random variables dened on a probability space (, F, P).
{Y
t
(), , t T }, such that for each t T , y
t
() is a random
variable on the sample space , and for each , y
t
() is a
realization of the stochastic process on the index set T (that is an
ordered set of values, each corresponds to a value of the index set).
Time Series: A time series is a set of observations {y
t
, t T
0
}, each
one recorded at a specied time t. The time series is a part of a
realization of a stochastic process, {Y
t
, t T } where T T
0
. An
innite series {y
t
}

t=
Eduardo Rossi c - Time series econometrics 2011 2
Mean, Variance and Autocovariance
The unconditional mean

t
= E[Y
t
] =
_
Y
t
f(Y
t
)dY
t
(1)
Autocovariance function: The joint distribution of
(Y
t
, Y
t1
, . . . , Y
th
)
is usually characterized by the autocovariance function:

t
(h) = Cov(Y
t
, Y
th
)
= E[(Y
t

t
)(Y
th

th
)]
=
_
. . .
_
(Y
t

t
)(Y
th

th
)f(Y
t
, . . . , Y
th
)dY
t
. . . dY
th
The autocorrelation function

t
(h) =

t
(h)
_

t
(0)
th
(0)
Eduardo Rossi c - Time series econometrics 2011 3
Stationarity
Weak (Covariance) Stationarity: The process Y
t
is said to be weakly
stationary or covariance stationary if the second moments of the
process are time invariant:
E[Y
t
] = < t
E[(Y
t
)(Y
th
)] = (h) < t, h
Stationarity implies
t
(h) =
t
(h) = (h).
Eduardo Rossi c - Time series econometrics 2011 4
Stationarity
Strict Stationarity The process is said to be strictly stationary if for
any values of h
1
, h
2
, . . . , h
n
the joint distribution of
(Y
t
, Y
t+h
1
, . . . , Y
t+h
n
) depends only on the intervals h
1
, h
2
, . . . , h
n
but
not on the date t itself:
f(Y
t
, Y
t+h
1
, . . . , Y
t+h
n
) = f(Y

, Y
+h
1
, . . . , Y
+h
n
) t, h (2)
Strict stationarity implies that all existing moments are time
invariant.
Gaussian Process The process Y
t
is said to be Gaussian if the joint
density of (Y
t
, Y
t+h
1
, . . . , Y
t+h
n
), f(Y
t
, Y
t+h
1
, . . . , Y
t+h
n
), is Gaussian
for any h
1
, h
2
, . . . , h
n
.
Eduardo Rossi c - Time series econometrics 2011 5
Ergodicity
The statistical ergodicity theorem concerns what information can be
derived from an average over time about the common average at each
point of time.
Note that the WLLN does not apply as the observed time series
represents just one realization of the stochastic process.
Ergodic for the mean. Let {Y
t
(), , t T } be a weakly
stationary process, such that E[Y
t
()] = < and
E[(Y
t
() )
2
] =
2
< t. Let y
T
= T
1

T
t=1
Y
t
be the time
average. If y
T
converges in probability to as T , Y
t
is said to
be ergodic for the mean.
Eduardo Rossi c - Time series econometrics 2011 6
Ergodicity
To be ergodic the memory of a stochastic process should fade in the
sense that the covariance between increasingly distant observations
converges to zero suciently rapidly.
For stationary process it can be shown that absolutely summable
autocovariances, i.e.

h=0
|(h)| < , are sucient to ensure
ergodicity.
Ergodic for the second moments
(h) = (T h)
1
T

t=h+1
(Y
t
)(Y
th
)
P
(h) (3)
Ergodicity focus on asymptotic independence, while stationarity on
the time-invariance of the process.
Eduardo Rossi c - Time series econometrics 2011 7
Example
Consider the stochastic process {Y
t
} dened by
Y
t
=
_
_
_
u
0
t = 0 with u
0
N(0,
2
)
Y
t1
t > 0
(4)
Then {Y
t
} is strictly stationary but not ergodic.
Proof Obviously we have that Y
t
= u
0
for all t 0. Stationarity
follows from:
E[Y
t
] = E[u
0
] = 0
E[Y
2
t
] = E[u
2
0
] =
2
E[Y
t
Y
t1
] = E[u
2
0
] =
2
(5)
Eduardo Rossi c - Time series econometrics 2011 8
Example
Thus we have = 0, (h) =
2
(h) = 1 are time invariant.
Ergodicity for the mean requires:
y
T
= T
1
T1

t=o
Y
t
= T
1
(Tu
0
)
y
T
= u
0
Eduardo Rossi c - Time series econometrics 2011 9
Random walk
Y
t
= Y
t1
+u
t
where u
t
WN(0,
2
). By recursive substitution,
Y
t
= Y
t1
+u
t
Y
t
= Y
t2
+u
t1
+u
t
.
.
.
= Y
0
+u
t
+u
t1
+u
t2
+. . . + u
1
The mean is time-invariant:
= E[Y
t
] = E
_
Y
0
+
t

s=1
u
s
_
= Y
0
+
t

s=1
E[u
s
] = 0. (6)
Eduardo Rossi c - Time series econometrics 2011 10
Random walk
But the second moments are diverging. The variance is given by:

t
(0) = E[Y
2
t
] = E
_
_
_
Y
0
+
t

s=1
u
s
_
2
_
_
= E
_
_
_
t

s=1
u
s
_
2
_
_
= E
_
t

s=1
t

k=1
u
s
u
k
_
= E
_
t

s=1
u
2
s
+
t

s=1
t

k=1
u
s
u
k
_
=
t

s=1
E[u
2
s
] +
t

s=1
t

k=1,k=s
E[u
s
u
k
] =
t

s=1

2
= t
2
.
Eduardo Rossi c - Time series econometrics 2011 11
Random walk
The autocovariances are:

t
(h) = E[Y
t
Y
th
] = E
__
Y
0
+
t

s=1
u
s
__
Y
0
+
th

k=1
u
k
__
= E
_
t

s=1
u
s
_
th

k=1
u
k
__
=
th

k=1
E[u
2
k
]
=
th

k=1

2
= (t h)
2
h > 0. (7)
Finally, the autocorrelation function
t
(h) for h > 0 is given by:

2
t
(h) =

2
t
(h)

t
(0)
th
(0)
=
[(t h)
2
]
2
[t
2
][(t h)
2
]
= 1
h
t
h > 0 (8)
Eduardo Rossi c - Time series econometrics 2011 12
White Noise Process
A white-noise process is a weakly stationary process which has zero
mean and is uncorrelated over time:
u
t
WN(0,
2
) (9)
Thus u
t
is WN process t T :
E[u
t
] = 0
E[u
2
t
] =
2
<
E[u
t
u
th
] = 0 h = 0, t h T (10)
If the assumption of a constant variance is relaxed to E[u
2
t
] < ,
sometimes u
t
is called a weak WN process.
If the white-noise process is normally distributed it is called a
Gaussian white-noise process:
u
t
NID(0,
2
). (11)
Eduardo Rossi c - Time series econometrics 2011 13
White Noise Process
The assumption of normality implies strict stationarity and serial
independence (unpredictability). A generalization of the NID is the
IID process with constant, but unspecied higher moments.
A process u
t
with independent, identically distributed variates is
denoted IID:
u
t
IID(0,
2
) (12)
Eduardo Rossi c - Time series econometrics 2011 14
Martingale
Martingale The stochastic process x
t
is said to be martingale with
respect to an information set (-eld), I
t1
, of data realized by time
t 1 if
E[|x
t
|] <
E[x
t
|I
t1
] = x
t1
a.s. (13)
Martingale Dierence Sequence The process u
t
= x
t
x
t1
with
E[|u
t
|] < and E[u
t
|I
t1
] = 0 for all t is called a martingale
dierence sequence, MDS.
Eduardo Rossi c - Time series econometrics 2011 15
Martingale difference
Let {x
t
} be an m.d. sequence and let g
t1
= (x
t1
, x
t2
, . . .) be any
measurable, integrable function of the lagged values of the sequence.
Then x
t
g
t1
is a m.d., and
Cov[g
t1
, x
t
] = E[g
t1
x
t
] E[g
t1
]E[x
t
] = 0
This implies that, putting g
t1
= x
tj
j > 0
Cov[x
t
, x
tj
] = 0
M.d. property implies uncorrelatedness of the sequence.
White Noise processes may not be m.d. because the conditional
expectation of an uncorrelated process can be a nonlinear.
An example is the bilinear model (Granger and Newbold, 1986)
x
t
= x
t2

t1
+
t
,
t
i.i.d(0, 1)
The model is w.n. when 0 < <
1

2
.
Eduardo Rossi c - Time series econometrics 2011 16
Martingale difference
The conditional expectations are the following nonlinear functions:
E[x
t
|I
t1
] =

i=1
()
i+1
x
ti
_
_
i+1

j=2
x
tj
_
_
Uncorrelated, zero mean processes:
White Noise Processes
1. IID processes
(a) Gaussian White Noise processes
Martingale Dierence Processes
Eduardo Rossi c - Time series econometrics 2011 17
Innovation
Innovation An innovation {u
t
} against an information set I
t1
is a
process whose density f(u
t
|I
t1
) does not depend on I
t1
.
Mean Innovation {u
t
} is a mean innovation with respect to an
information set I
t1
if E[u
t
|I
t1
] = 0
Eduardo Rossi c - Time series econometrics 2011 18
The Wold Decomposition
If the zero-mean process Y
t
is wide sense stationary (implying
E(Y
2
t
) < ) it has the representation
Y
t
=

j=0

tj
+v
t
where
0
= 1,

j=0

2
j
<

t
WN(0,
2
)
E(v
t

tj
) = 0, j
and there exist constants
0
,
1
,
2
, . . . , such that
V ar(

j=0

j
v
t
) = 0.
Eduardo Rossi c - Time series econometrics 2011 19
The Wold Decomposition
The distribution of v
t
is singular, since
v
t
=

j=1
(
j
/
0
)v
tj
with probability 1, and hence is perfectly predictable one-step ahead.
(Deterministic process).
If v
t
= 0, Y
t
is called a purely non-deterministic process.
Eduardo Rossi c - Time series econometrics 2011 20
Box-Jenkins approach to modelling time series
Approximate the innite lag polynomial with the ratio of two
nite-order polynomials (L) and (L):
(L) =

j=0

j
L
j

=
(L)
(L)
=
1 +
1
L +
2
L
2
+. . . +
q
L
q
1
1
L
2
L
2
. . .
p
L
p
Eduardo Rossi c - Time series econometrics 2011 21
Box-Jenkins approach to modelling time series
Time Series Models
p q Model Type
p > 0 q = 0 (L)Y
t
=
t
AR(P)
p = 0 q > 0 Y
t
= (L)
t
MA(Q)
p > 0 q > 0 (L)Y
t
= (L)
t
ARMA(p,q)
Eduardo Rossi c - Time series econometrics 2011 22