Documente Academic
Documente Profesional
Documente Cultură
Contributions
III.
By Karl
to the
Received October
18,
"Plates
1893,
5.]
Contents.
Page.
I.
71-85
9-10
II.
85-90
On the Dissection
of
Symmetrical Frequency-Curves.
III.
measurements
Table
I.
Table
II.
First Six
Carapace of Prawns.
Powers
16-18
100-106
Numbers
106
I.
of the
90-100
On
107
10, 1894
the Dissection
107-110
of Asymmetrical Frequency-Curves.
The centre
or
origin for
frequency-curve
may
all
the
we
if
may
mean
is
of that range.
'"
# Symmetrical shapes
Part
II.,
may
of course occur
See
11 of this paper.
9.5.94
PROF. K. PEARSON
72
of measurements
series
gives
something approaching a
impartially round
economic
the mean.
measurements
rise
it
to a normal
condition
stable
termed a normal
is
there
biological,
however, a well-marked
is,
curve,
there
When
curve.
sociological,
deviation
from
and
this
arise
curves having parallel but not necessarily coincident axes and different parameters.
Even where the material is really homogeneous, but gives an abnormal frequency-curve
the amount and direction of the abnormality will be indicated if this frequency-curve
can be split up into normal curves. The object of the present paper is to discuss the
The equations for the
dissection of abnormal frequency-curves into normal curves,
dissection of a frequency-curve into n normal curves can be written down in the same
manner as for the special case of n == 2 treated in this paper they require us only to
But the analytical difficulties, even for the case of n = 2,
calculate higher moments.
;
it
may be
why
first
case.
into
two
species, rather
is
of special importance.
we
in attempt-
Lastly, even
where the heterogeneity may be threefold or more, the dissection, into two is likely
In the case of
to give us, at any rate, an approximation to the two chief groups.
homogeneous material, with an abnormal frequency-curve, dissection
curves will generally give us the
much, then,
may
amount and
into
two normal
So
must be made between the two cases which may theoretically occur.
If we have a real mixture of two normal groups represented by our abnormal frequencycurve, then, theoretically, it is possible to find the two components, and these two
components must be unique, If they were not unique, a relation of the following kind
must hold for every value of x :
distinction
e
<r
lX/(27r)
20Y*
_
_|_
o-
-*J*2or
2^8
e
'C\
>/(27r)
/-f- ^
lyo
a?jX
(2tt)
'J
2v.J>
_L
lit
t^
o 4V/ (2tt)
PBOF.
K.
PEABSON
Between the
ON*
an
infinite variety of
by the same
s>
~L-j
1
a*!
The
jl
first
<r3
cr
is
or
For
(*-fti) 2
~*
= -A 6
*iV
^2
an(^ putting
l<x{"
<r L
*J
~|
P.
W~<riV
<T A
*s
unless
seems
it
o- x
ia
Dividing by -jj^~\
&& anc 0V
73
limit,
cr4 .
c3
true,
say
must be
it,
we
either side.
2o-/
C4
ultimately, to c 3
Our
=c
e""
25
?,
same manner
may now
as before to
cr2
<x4 ,
and,
be written
(x-b 3 V
<r
4.
original equation
<
x2
2or4
CT,
CTc
Dividing again by
= _*
20"! 2
_>-&
'2
2cr,
(3
(3
lV / (2tt)
^sV
h=
2)
2<tf
( 2?r )
fa)-
or
h+h=
Similarly, putting
-J
(6 2
&*)>
we
b2
64
61
=&
&3
=h+
3>
or
h+
MDCCCXCIV.
A.
()
6,
"(a) is true,
THEORY OF EVOLUTION.
74
Multiply equation
(tj)
above by
x,
x and x B
?J
(6 l
{So-!
- h) + 6 -
(&!
reducing by aid of
ft}
and
(a)
63) c
in succession,
We
a #
c
= (h
=
*i
results
find
/>
4:
c3
2)
{3a-/ (6 4
68 )
+ 6 - hi)
4
-.
(/3) ?
c 2f
to
(/3)
and
3)
reducing by aid of
80-/
36/
8<r 2
3/>
(a),
determine b l9
h%,
bg
&4
and
(8),
(/3),
,
we must have
cr 3 ,
<?
=5
61
and
Now
c2
62
sum
4& x & 3
4& 2 &j,
3&/
<x l3 cr 2 ,
hence
64
in one
sum
it
of
it
is
We
must
c 1?
c.
3>
(8),
.....
and
c2
But
(e).
5 1? 5 2 are in
is,
up into
abnormal frequency-curve
;
indeed, if
it
will
be not a mixture,
asymmetrical binomiaht
bi}
That
of two normal-curves
2
2
it is
and
l5
8)
3
-f 3/> 3
as definite functions of
be true, or
(e)
two forms,
(y) to the
2o-/
2a"!
Equations
and
(a), (/3),
g)
it
may
the limit to an
be
by any method of dissection. A mathematical criterion for dismight easily be given. For example, in the method
of the present paper, we might define that as the " true," or at any rate the " best,"
solution which gave for the compound-curve a sixth moment nearest in value to that
Such a theoretical criterion, however, may not have much
of the observation-curve.
solution be given
The values
of the successive
moments
of
5 of this paper,
and
where
0,
c,
and
slightly fuller
ft
form
is
to
is
'
/3
is
always positive.
[A
PROF. K. PEARSON
practical value.
moments
be
will in general
if
moment
on other organs-suffices
in
of
Added
e.g.,
is
two
much
to this
sufficient to deter
measurement
results of
found.
moments
first five
75
we have
solution
different
methods of
whether we
or error of
mean
normal curve to a
fitting a
square,
to be content with a
mean
depending on
"
or the median, and proceed by
quartiles," mean error
series of observations,
in
some cases
to slightly different
As from the
utilitarian standpoint
first
good results
may
for a simple
error of
mean square from the second moment, so it seems likely that the present investigation,
based on the first five or six moments of the frequency-curve, may also lead to good
results. While a method of equating chosen ordinates of the given curve and those of
the components leaves each equation based only on the measurements of organs of one
size,
all
it
will, thus,
to discover
two
(2.)
Any
method of moments
to the
figs.
and
2,
In the case of a frequency-curve whose components are two normal curves, the
complete solution depends in the method adopted in finding the roots of a numerical
equation of the ninth order.
many
(i.)
its area.
Mean
trials
and
may be
the position of
Mean
Square"), and
its
component
Innumerable relations
Clearly
found, but
failures.
them lead
to
PROF. K. PEARSON
76
exponential equations, the solution of which seems more beyond the wit of
man than
In any given example the conditions will be sufficient to reduce the suitable
(3.)
two or even
one.
These limiting
This quadratic
two
cases
its
We
have
two
funda-
Both
(i.)
is
species, each of
which
may be
is
is
the
sum
two normal
of
curves.
The
The
process of differentiation
One
(ii.)
root is positive
and
now
The
birthupon
as
the
looked
now
be
may
population (unselectively diminished by death).
The negative probability-curve is a
selective diminution of units about a certain mean
that mean may, perhaps, be the
is
possibly
possible that in
is
w ill be found
r
teristic
if
to exists
some numerical
(i.)
and
two types do
exist,
(ii.)
then one will give a much better agreement with the actual
elimination of
(iii.)
members round a
selected mean.
occur.
We now
to
break
it
(4.)
(i.)
line
y y\
if
we take
the
sum
of the products of
every element of area by the wth power of the distance of the element from the line
y'y'\
n th
y'y' 9
then the
A to C in
our
y dx, the integration extending all over ABC,
where the curve is always bounded by a straight line, AC, perpendicular to yy.
moment
case,
n= \x
or from
order h a
if
a be the area
We
numerical factor.
of ABC.
moment
where
/i n is
is
77
of the
a purely
factors.
Given the
first
/i
[/%,
/x'
3,
// 4
\i n ,
fl tt
h H 0L
'x
d) n y dx,
or
U (n
Mi
Mi
Ms
m's
12
7 (?&
'
A1
^-s
M5
When
the line
symmetrical about
the right of
yy,
or
-2)
1,
Ms
M4
1)0*
+ 6gV8
5<?m'i + lOgVs
m's
yY
yy
d
"
42m's
m'*
+2
lOgVg + 5<2Vi -T
42
-
0).
m'i
is
we take yy
to
\x!
x,
ju/
3,
negative,
ytt'
are
all zero.
Hence
if in this
case
!?8
/*l
<1
H = P* + f
H = 3 2^2 + <f
l
Pi
= H-\ +
(2)-
6<zVa
</'
The distance of the centroid of ABC from yy is the ratio of its first moment
fi\ha to its area a, and = [i-Ji.
(iv.) To find the successive moments of a given curve about a given line.
For the purposes of the present problem we require only the first five moments of
(iii.)
a curve like
ABC
The
its centroid.
solution
may be
Analytically.
equal change in
individual
the
deviation.
point
corresponding to an
beyond
C,
which
no
1000, and h
let
be y l9 y 2 yZ9 y^ y 5
x y
may
1"
X2/ 3
y3
y4
be conveniently taken
100,
^xy^
x i,
x yx
2-
x y%
know
the
occurs,
deviations before
or since a
AC
results of
jx/
.,
3,
jx/
4 , fi
&9
provided we
the 2nd, 3rd, 4th, and 5th powers of the natural numbers up to about 20 to 30.
THEORY OF EVOLUTION.
79
^2
+ 6gV' - 3?*
5g/*' 4 + 10? V' lOg^/Zj +
4^'3
^4
fa
rs
HUMUf
g J
(4
4g
5
*P*r
The centroid having been found, it may be asked Why we should not calculate
The answer lies in the fact that the centroid will not generally
fa, fa, fa, fa directly ?
coincide with a unit division on the deviation axis, and the powers to be calculated,
instead of being those of two place figures, become in general powers of numbers
containing three or four figures.
Thus the labour of the arithmetic is much increased.
(b.) Graphically. If the figure be drawn on a large scale, the moments may he
found with a fair degree of accuracy by aid of the following process, which has. long
been of use in graphical statics for finding the first, second, and third moments, of
:
plane areas.*.
It
is
straight line
PP
first to
OCA.
Take O'y"
O'y from
AC
to
parallel to O'y
ABC]
ABC, bounded by
and at distance
h.
Take any
the
line
it
let the
it
let the
perpendicular from
it
let
in N',
in
moments about
Q2
on O'y" meet
in
are determined.
The
third
moment
of a plane area
is
been long in use to find " equivalent figures " in the case of beam
sections.
PROF. K. PEARSON"
80
^i* ^2*
AQ L C AQ
curves.
^3
J.
line
off
(4)
/V
= a i/0
^/
a6 / a
about a
L 2 C, &c, be read
iien
may
if
on a
Or, the
we may
first
shift O'y
till it
pieces of their areas negative, to carry the planimeter point, in the proper sense, round
their contours.
(5.)
To
c is the
of y\
Let them be
'
and
'.
Then
a
M/=2
M/ =
Clearly
[*
My
'
and
'
yx dx
=c
o-
yx 'dx
=c
3cr
9,
*a
2
J
2
.
are zero.
The method
is sufficiently
them
first
four
moments
of a
number
by him. February
9,
1894.]
Now
(ii.)
let
81
Let us use
IBr
Equations (2) of Art. 4
taking
(ii.),
as the axis of
yy
left,
symmetry of the
probability-
then
= be.
flJlO
+b
(36o- + V)
(3o- + 66V + 6)
(15cr 6+ 106V + 6 )c.
(cr
) C.
c.
2
4,
c.
Now
let
cJol
alb
z,
= u,
and
b/'h
= y.
M
M
M
M
M
= yzah,
==y%(l+^)aA3
(6.)
We
are
now
(^).
in a position to write
= yh(l + 3u*)ah\
= y% (1 + 6w + 3tt aA
= y 2 (1 + 10^+1 5^ oA
s
and we have
down the
let
Let
this centroid
and the
first five
Then,
by the
if
/x,
5,
momentbe found
(iv.).
quantities b l9 c l9
cr }
and
&2 c 2
,
7v
we
/x 4 , /x
have, since
MDCCCXCIV.
A.
z i>
<x 2 ,
u i>
Z9
i
82
+ c = a,
(Yl z + 7&) ah = ,
{y \ (1 + Mj + y \ (1 + w )} aW = ^ a/i
(! + 3m
+ y 2 (i + 3% )} ^ = A
{ri
{y^ (1 + 6< + 3<) + yA (1 + 6% + 3%*)} '** = /*M*.
(1 +
{yi
V + 15 V) + y/% (1 + 10 + 15?/,/)} A =
C]
2
2
3z i
fi 8
The
its
first
/x
a/*
6
.
equation here represents the equality of the areas of the resultant curve and
components.
yj *! (1
y l \(l
5
y,
(1
yi%( 1
"*""*"'
es*09se
|
.1-.
......
+ V) + ya%(! + V) = ^a
+ 3<) + y % (1 + 3^) =
3
+ 6u + Su^) + y,\(l +
"
+ 10< +
six equations
15V)
/x
=^
15V)
(1-0).
-(H).
6ui+3u^)
+ jth (1 + 0% +
/p\
\J Fe
/*s
(12).
13 )-
Equations (8) (13) give the complete solution of the problem.* After several trials,
I find that the elimination of z v %, u l9 u 2 from these equations, and the determination
of equations giving y xy3 and y x
y.z appear to lead to a resulting equation of the
(7.)
Eliminating
z2
between
(8)
and
<V"|
Similarly
we have
(9),
"~
solution.
portions of area, or of
be found.
7i
7i
my
"*"
**
/*
* All
It is possible,
axis of
a?,
all
/,
r\
7s
less
complex kind
may
ultimately
PEOF. K. PEARSON
83
Equations (14) and (15) clearly give the numbers in the component groups so soon
as
determine
u^ and u^
in
and
zx
terms of y l9 y2
Solving them
z% in
we
=-
y Uf
x
to
find
(16).
7172
72
7173
These equations clearly give u x 2 and u%, and, therefore, the standard-deviations of
when y x and y3
are known.
V2
= Tl + 72.
2>L
(y*^)
jPa
Tir-2-
Then
2 (y x
-yV +
2
MsM
v,
^-
2)
Vc
^-
( yi
% - y %) + 3 - <) =
3
yiS
y*v = (yi y)
%
- yaX
(yi
- ys)
-y
t^ 2
(19):
2)
in (20)
and
V7l
"~*
72/
I
L
"72"
- Pi +
8
Ms
9P1
i'2
iJ 2
2^
-i> 3 -2/, 2
3
M2P1
9(^ 3
6p 2
_2
5/V^i
2
(21).
+ l^^
+ H
9
iJ 2
/**)
0.
y?2
2O/X3
(20),
tPiPa j>
whence,
a3
We find
(21).
72
^23
/V
-~ = (n- y )^-J
+ i^ + iPi
i/* 4/p 2 }
/^ te + P2
Pi
- *^^
+
/*2JPi
( 7l
7)
j/*a "~ 3
^-
|ip
2)
7i
+ P2
72
71
7i v i
iPl72
"J
we have
/*2
(18),
2p x
4 PlJp 2
i^2
15 (2^ 2 ^3
^3
aO
0.
^3
3 JP1P2
84
PROF. K.
Write
= 'W
\t
3 /^>
= 30 ^2/x
^5
3 /x 5
22 )>
6P2 8
24 )>
and put
^ ~ ^Hlh
5/x 3
From
p3
- 2p + 4p
3
become
2^3 S
p2
- Va +
20/x 3
p2
- X p =0
first
find
p3
in terms of p%.
+ Ps (W +
hPz
2p 3
(25).
Multiply the
by p 3 and
first
HPz
Multiply (24) by
W+
W+
(-
Ps
and add
2/x 3
to (26)
X4:p 2
we
2j9 3
3
)
20/iaPg
- X^* =
(26).
find
p -
2/x 3 X 4
X 5^2 3
X 5p/
8/x 3 p 2
0,
or
ql
^-
p
7
Eeturning to
containing
jpa
24^ -
28X4 p/
+ (288/*
4
3
~Pi7
is
36/is
-12X,X 5/x 3
and X 5
= 30/x
3//,/,
2 //, 3
while
V-
-X^
jp a ,
as
//,
3/x 5 ,
3
3
3
97 \
+ P2 =
.'
'
(28).
Ws - 10A/)
we
and
/x
see
we should
JPs
- (148^% +
2X6 );p a*
(29).
may always be
Thus
(8.)
^4 =
= p /p
only, on
This equation
g^S^jPg
this equation.
found.
0, all
is
of an odd
9/x 2
for
3/x 4
which
first.
it
will be needful to
work
their
values out to a greater degree of exactness the slighter the degree of asymmetry.
PROF. K. PEARSON
We may
Stage
is
and
curves, equations
L Find
85
/x
3,
/x
3,
/x
4,
/x
3,
X4
and X v
Stage
Stage
from
Solve (29) for y and find the corresponding values of p from (27).
III Find the positions of the axes of the component normal curves
II.-
(28).
Stage IV.-
The
and
fractions z Y
Pi
Stage V.
Since
crjh
= */v
and crjh
and
(30),
= \/%
Pz
z*
4p2
'
(19).
(9.)
Abscissae.
Ordinates.
2
3
5
2
5
6
10
13
19
20
25
40
31
60
62
54
10
11
12
13
14
15
Abscissas.
16
17
18
19
20
21
22
23
24
25
26
27
28
29
'1
Ordinates.
74
84
86
96
85
75
47
43
24
19
9
5
1
millim.
86
This curve
is
1,
fig.
1,
and
is
clearly asymmetrical.
:-
were obtained
=
=
^=
=
=
16799
fL{
304-923
'
/* B
/x
116,061*435
'
2,385,609-719
ix 5
fii,
since h
1, is
5,831759
'
i.e.
= q of p.
77
(ii.).
(1), p. 77.
=
=Pi =
1576-533,413
85'205 407
7920'604 761
22-716,599
1jl.
53-874,770
ix s
X5
by
(22) of p 84.
(29), let it
the form
p*
Then the
coefficients a s
a3
a3
a4
a6
a7
a8
a9
=
=
=
a 6
o.
99-406
4,353-742
423,696
=
=
=
=
t5
t/
/ \J
uy) DO
jZii/Oji/
X1
1,232,409,400
957,080,900
24,451,990,000
PROP. K. PEARSON
87
pz =
9
10x and divide by 10 we then have for the fundamental nonic the
following equation, where only three decimal places are retained
Put
"
994X7
fi
119-299x
123-241X
0.
./(/l
/.(
oo
=+
oo
4-
oo
A
f.
/
/
/
/
fl
co )
oo )
7 ( )
8 (
9 (
<X)
CO
=+
=+
=+
= =:
)
)
3=
/.()
/
/ (_Q0)
/ ()
7
3 real roots.
Two
roots
One
root between
Xi
Xa
Xz
With
1,
-5,
'
sufficient accuracy
and
between
and
To
=+
==+
=+
6 changes.
As
=+
/9(
3 changes.
-65,
3rd
-870,
-670,
422,
-8757,
-6724,
4170.
= 8*757.
i>a= 6724.
4-170,
_p =
1st solution, jp 3
Xi an d Xa*
1, Xs
-89,
2nd
1,
40,
5,
and gave
88
p. 84,
and then^j
The quadratic
= p /p
3
p2
There resulted
found.
p 3 was
'8*757.
first
1*027.
frequency-curve),
is
:
y
whence
y,
V027y
8757
3-517,
0,
y3
2-490.
zl
%=
-4145,
p. 82.
'5855,
ex
414*5,
c2
585*5.
The values of the standard-deviations, or l and cr2 were now determined from
3
At the same time the
(18) and (19), where, since h = 1, v x = cr^, and % = cr3
maximum ordinates of the component probability-curves, y l and y25 were found from
,
ci
Vl
- ^(27T)
%
v/(27r)
O",
Org
There resulted
o-!
y1
Thus the
1st solution
1st
cr
2
uo
-*
as follows
= 4*4685,
= 37-008,
may
be
summed up
Component.
=
by =
=
yi =
Cr
o-y
3*1154.
4fc
Jomponent.
^jllCt v.
414-5,
ca
3-517,
63
4-4685,
er 2
37-008,
2fe
= 585-5.
= 2-490.
= 3-1154.
= 74-976.
These two normal curves were now drawn by aid of the Table
calculated afresh for this purpose from the exponential/*
in
fig. 1,
remarkably
close
it
close as
more convenient
to
we
which was
this curve is in
II.,
is
usually tabulated.
PROF. K.
number
fact
the observation-curve
that
89
resultant.
2nd
Solution.
p =
2
Precisely
and
6*724,
similar
undertaken
for
the value
will,
it
were
calculations
here.
Quadratic for y
1st
Ci
61
cr l
*3412y
0.
Component.
=
=
=
=
2nd Component.
= 532*8.
2*428.
b =
=
4*7702.
44*559.
ya =
467*2,
c2
2*769,
2*878,
cr
64*764,
We
6*724
in fig.
2,
3rd Solution
While the
(p 2
=yy
{
same
i2
p%
first
of the
4*170.
same
this solution
makes y and y 2
{
component curves
fall
both on the
one of them must be negative, and the solution promised to be a subtractive one,
to represent the frequency-curve as the difference of
Determining
p3
The
(27),
we
3'605y
find
two normal
^=
4*170
we cannot
difference of
So
(10.)
biologist's
curves.
;
hence
0.
therefore,
3*605
work
is
now
accomplished.
Taking the
unequal groups about two different means, then one or other of our solutions
correct answer.
/jl
192,446.
i.e.,
first
According to this
test,
solution
the
it is
first
188,099 and
solution
is
for
is
two
the
the required
one,*" but, as
we
have noticed, the two solutions are themselves much closer together than either to
*
the
The theory
first
MDCCCXCI V.
A.
is
Professor
Weldon
tells
me
that
THEORY OF EVOLUTION.
90
In
75).
p.
fact,
and neither
differs
differ
statistical
measure-
this kind.
it is
up
Thus the problem indeed becomes somethe asymmetry of the frequency-curve becomes much
Professor
Weldon
differs
it
original obser-
mean according to
Without corre-
exact normal curves, our method gives two solutions and two only.
9
lated measurements,
it
might be
difficult to discriminate
moment
The perhaps
statistics.
over-fine theoretical
first solution.
Another important case of the dissection of a frequency-curve can arise, when the
two components,
i.e.,
is all
sum
then
consists of the
This case
are identical.
still
or difference of
are distributed
it is
not unlikely
he
is
is
dealing with an
by
its
and
fit
the
normal curves.
This, at
any
normal curve
would
r
of our material, although possibly our test for tw o
its
heterogeneity
in the case of
In
this case,
well.
sum
We
ought
first
two
stage to demonstrating the homogeneity
fit still
rate, is a first
fairly
is
better the
may
fail,
or difference of
is
*
multiple rather than double.
PROF. K. PEARSON
We
will
91
(= u^),
i\2
o-Jh
identically satisfied,
(=
In this case
we
if
unite,
we have
and
(8), (10),
-and (12)
&3
0,
(11)
(9),
v{
<rjh
and (13)
become
g^ I
.-j
bx
..........
JL
(ol j,
(8)-(13), put z1
= %,
Yi
*i
Yg>
We
= % and we have
U\
=H= i
(1
7i
+"i 2 )
= /^
7i
&<i
+ 3 V) =
/t
4,
whence
7l
^- ITCW^)
>
or,
cl
^2
j1
= -j = i/w-ftU
rt
.-{#"f1U^The
depends on
The following
an example
is
is
H\=
6-482
^3
of
ju,
indicates general
We have
g1
=c
and
symmetry
53
He
2-486
fired in
two groups
accordingly
= - & = 1-082,
= 1*1'^>
a =
1000 shots as a whole a = 1*577.]
3
For the
*2
would
almost exactly.
fy
divisions 5
^= -104
^= 15793.
The smallness
/*!=
/i
= 44-502
/ = 320-582
y = 2405-094
/*'
3/li^
and
arise
Mr. Mjerriman's
7,
and not at
from a change
figure,
that of 6.
'482,
we
find a
compound-curve
iu
92
PROF.
K.
<vi
^1^1
Clearly
we
require one
f .i-o^o
^i
%^2 ==
At
more equation.
.........
Wo
A1 4
3"
TILUO.UY OF EVOLUTION.
sight
first
'
it
is
the
sum
\^3).
equation would come readily, from the fact that the mid-ordinate
curve
'
(oZ),
of the frequency-
component probability-curves.
This leads to
LI+ /;oV" = m
/v
or
7;;+7k= M
'
<
34 >>
if
m =
mA/a.
\/(27r)
But besides the disadvantage of throwing our solution back on the correctness with
which we may have observed measurements of one size only, namely, the mean, the
result of eliminating between (31)-(34) leads to an equation of the eighth order. To
avoid this, it seems easier, as well as more accurate,* to take as the fourth equation
that obtained from the sixth moment.
Let n ah G be the sixth moment of the given frequency-curve about
symmetry, thent
C)
[jL
(
.ah G
Wi*
The solution of
Eliminating
have, writing
Zl
h
W
l
^=
and (35)
% we
K
K-
locr/V^
or,
=v
'%)
IV 2 )
is
2
,
l5or 2 c 2
(35).
easy.
w.2
= v/,
Hw
- /Xo^o,
=
*>
-l-fi,h
whence
__ T5
if;
j
-V
3 W>2
Mi - Ma'^s
Mfl
~ ^2^2
^2 ~ " 2
3 M-l
t Generally,
if
2r
be the 2r
moment
M =
2r
or,
2r
(2r
(2r
1) (2/-
I) ^M,,._,,
- 3)
axis of
T5
its
5.3.1ff 2r c.
PROF.
-E.
93
Thus
The two
w l and
complete solution
is
CiL
Ot
w w
l
where
wa
and
(i.)
/x
the
positive,
(iii.)
is
lie
/x
of
between
cx
l9
we have
< w2
positive, or
Obviously
zero,
w*
w and w2
Y
is
(^ 2 /x 4
|/x
6)
w -
(^ liwa)
"
86 )
if
and
cY
c2
is
is
curves.
positive
and
c 2 negative, or
c x is
negative and
we have
if
n\
two normal
>w
fx %
become
3^ 2
sum
(ii.)
Now we may
Let
Ow
are roots of
(fa
(12,)
Gn
^=
c 2 is positive, or
3/x 2
2
,
and
/x
5/a 2 jlc 4 ,
This gives
for all
satisfied if
all
the
is
If 4 be not equal to 3/x 2 and /x 6 be not equal to 5/x3 4 then we have no right to
assume that a symmetrical frequency-curve refers to homogeneous material. We must
then investigate whether a better result cannot be obtained by treating it as two
superposed normal curves having the same axis.
3
ju,
/z,
The quantities
_
I propose to call the excess
moment
/^j-3/V
and
,j
&
5 /^4
of the
fourth
moment over
moment
5 '"
may
the square of
moment from
Here
""-
defect
the second
one-fifth the
They appear
is to
to be a
good
94
PROF. K.
measure
We
may now
and
and
e^
e2 ,
defect.
w
W
%
+
fa/
fa
X
3ci(
2,
3 l (l
e2
el )
= 0.
This gives
_____
e3
^/{(e z
6 ei f
36e_ 8 }
^2
We have
x
(i.)
also be
and
e2
both positive.
both positive,
that
fore,
of
e2
(ii.)
ex
and
e2
both negative.
<
3e 1 (1
Then w
Cl ).
be real
will
if,
when
v/(-*i)<l>
(
2)
does not
lie
between
6(-e
){l
+ /(-
v
)}
and
6(-i)
(l-y(-
ei )}.
If
then
we must have
w may
+ a/(-i)}-
{1
ei )
positive,
we must have
or
(-e a )>3(-e
This latter condition
is
){l
if
-(-<*)}.
clearly satisfied if
v/(-ej)
On
6e l
>
J.
95
it is
(-0(1- (-e,)}
than
is less
Hence, our
\/(-i)> h
then
but
+ y(-e )};
1
if
then either
(or
it
must
lie
6g
)>6(- ){l +
1
v/(-i)}.
between
3
(-0(1
-(-e,)}
and
6(-e
X
(iii.)
positive
and
negative;
e2
){l
if
v/ (-^)}-
the values of
must be negative,
the values of
%o
must be negative,
ex
negative and
e.
positive
if
Thus we conclude
If the excess and defect are not zero, the frequency-curve, although symmetrical, is
not normal. If the excess and defect are of opposite signs, then the frequency-curve
cannot be broken up into the sum or difference of two. normal curves with common
The frequency-curve, if compounded of normal-curves at all, is of a higher and,
axis.
more complex character. If the excess and defect are of the same sign, then,
provided certain relations hold between the numerical values of the excess and defect
:
given in
(i.)
and
(ii.)
above, there
is
w hich
T
resolves the
example.
details
crabs
at
Professor
Weldon
discussion
by the consideration
my
of a numerical
by placing the following statistics for a shell measurement No. 4 of his series
my disposal. The resultant-curve and the corresponding normal curve are
pictured in
fig.
3 (Plate
3).
we
could not expect a more symmetrical result, or a closer graphical agreement, with the
normal curve.
as
we
But
is
this a real or
96
PROF, K.
Crab Measurements.
No.
Number
(Total
4.
_.,..._,..-.
of Crabs
~
'".
.
Ordi nates
xv 0dC1SSo6
(1 unit
5
11
4
5
6
40
98
121
152
147
55-
8
9
10
The
first six
of Table
I.,
moments were
except that a
h equals unity as
Pi
P$
20
now
we go a
Pl
[l
Pg,
=
=
by aid
/x,
Pf,
= y^ = 27403.
7-5092
==
Pi
Pa ~
2,150,845-6867
<r
9,
have
101-3022
1 crab).
13,334-0710
"
72
41
28
8
165,488-8438
p6
126
82
1,129-9971
(i unit
11
12
13
14
15
16
17
18
19
9-684,684
"
We
before.
999.)
"
Ordinates
Abscissae.
1 crab).
=
'
3-4751
176-7280
271-6007
7,919-2781
= /x/ =
/*4
"~"
3/^2
7*5637
044 712
?
10/x^ 8
22*6911
10*6485
X5
l
31*9455
1283*8486
*606 45
62
If
zero.
we had
a perfect probability-curve,
This, of course,
/x 3 , /x
comparative smallness of
/x
3,
ju,
5,
X4 X 6
,
e3 ,
5,
in
and
/x
3/x/,
any actual
es
and
/x
5/x^ 4 ought to be
set of observations,
shows a very
fair
but the
approximation to
PROF.
Since
e3
K.
>
3c! (1
and accordingly
e x ),
we
possible to break
it is
97
w we
have
^= 3-50971,
whence from
p.
93
1-01148,
=a
Cl
W%
0-!==
-0046,
v^ X
c2
3-50971),'
o~ 2
=aX
=y
7
1*0046,
(/x
1*01148)
or
cx
o-j
For
all
= - 5, #
= 5*134,
c2
cr 2
= 1004,
= 2756.
(c
999,
we have used
(14.) Hitherto
original group.
heads," then
in all
is
we
same groups as
to practically the
homogeneous
may be
What we
in
first
species.
readily tested
investigation would
13
is
fail,
/x
X 4 and X5 already
a
The nonic being
values of
Pi
by adding together
or totals.
Should these
is
improbable a priori.
given.
^ 7iV
+ aPa +
The nearest whole number is here taken for the Crabs in each group,
t The arithmetic throughout was of course of a most laborious character.
MDCCCXCIV.
A,
0>
98
a3
18-11448.
=
=
=
=
=
-
<%
Cl~
a6
7
a8
ag
Writing
p.2
x>
we have
f r
325-54964639.
1604-777825,114,
977-342,6614.
3154-2006888.
4412-284,2437.
1761-180374.
its first
derived function*
_/j
(x)
/( x )
=x +
18-114,48x
+ 325-549,646x 1604-777,825^
+ 977-342,661x + 3154-200,689x'
+ 1761-180,374 = 0,
9
26-472,95x
5
3
4412-284,244x
and
= X + 20-590,07 X - 12-076,32 X
+ 180-860,915x 713-234 589x
+ 325-780,887x + 700-933,486x
490-253,805.
5
/i (X)
The Sturm's
functions were
/(x)=
fM=
f(x) =
/(x)=
/*(x)=
now
Mx) =
fs (x) =
fa (x)
Totals
X=
+
+
co
o.
Mx)=
fa (x)
+
+
4-
+
+
+
miJn"ia
-j
-.
change
4 changes
5 changes.
00
9<
roots between
one.
PROF. K. PEARSON
This root was
now
99
Putting p 3
to/x' ]n ^ ie original nonic, I easily
and 1, then between '15 and '16, and by a succession of
localized.
finally '15326.
Thus
p%
1-5326.
p s was
= pjpo,
was
/
which has both
its
is
1-5326
0,
roots imaginary.
Thus, considerably to
that there
2'17245y
my surprise,
but greatly to
my satisfaction, it
was demonstrated
The
material
is
to
Professor
Weldon's
families
having been
dredged up.
On
the other hand, I cannot think that for the problem of evolution the dissection
of the
will
is,
results give
an asymmetrical curve,
Stage
(ii).
is
if
moment
will,
perhaps,
Endeavour
to break
if it
cannot be
broken up, either into normal components with non-coincident axes or normal components with coincident axes, the material
is
and we
shall
we
if
the numbers in
its
If the
numbers
* Beetillon has found a double-humped frequency- curve for the height of the inhabitants of the
o 2
100
These stages seem to represent the mathematical treatment of this portion of the
problem of evolution.
(15.)
on tracing
I found
root,
p3
minimum
'82 it reached a
< 1254
its
at
to be
'8497.
and
between
_p 2
2.
tendency
" in
This led to
p
y
2,
that near
any w ay corresponded
7
5*2521y
'8497
y2
5-40915.
to
I therefore investigated
first
it
"
at
4,"
'82,
and found
it
5*2521, whence
0,
or
7l
Whence
-15705,
nearly
zl
or the
numbers
-972,
-028,
c2
in the
cx
971
and
"tendency to a root"
in
28.
no way
fits
"forehead" case, and No. 4 measurements neither break up, nor have they even
a tendency to break up, in the same manner as the " foreheads." Since the nonic
between
may
*9
which
and
may
it
1,
is
and
to
two
we may note
is
that the
a minimum,
it
lies
We
III.
(16.)
for
it
The
method
might
when
illustrated.
Mr. Bateson
of
to each
as likely to arise,
The
when
101
a considerable
incidents of this case will be generally true for other groups of biological measure-
such groups of
mathematically analyzed.
In order to throw more light on the general question, I have added the following
analysis for the case of Prawns, the measurements for which were kindly placed at
my
disposal
Palcemon serratus.
Measurements
998
in
Measurements reduced
to thousandths of body
Number of specimens.
length.
3
3
4
11
24
38
56
80
105
121
117
108
77
69
62
48
may
Number of specimens.
length.
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
and
Measurements reduced
to thousandths of body
is
49
50
51
52
25
17
11
8
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
1
1
as
(17.)
The curve
(see
fig.
to me,
which
fundamental nonic.
4) given
will
be at once seen to
PROF. K. PEARSON
102
OIST
be distinctly asymmetrical.
coordinates, and using the
= d (=
=
fi\ =
fi\
same notation
q)
4,963-876,753,5
;jl\=
94,386*734,469
1,920,725-520,040
X4
X5
=
=
14116,678,13
fi 3
33*424,02673
/x
of
1,288*640,094,26
== 16,752*563,9961
2072-394,903
36,102-605,1706.
as a whole
errort obtained
fi 2
/x
o-=
The mean
we have
^=
=
=
=
16-191,382,8
276-277,555
fi/ s
//5
as before,
origin
given by
= \/^
or
3*7572.
from
<x
directly
is
=2*9978
2*8776.
directly 4*4087.
first moments, is a very good test of the asymmetry of the frequencyIn the very symmetrical measurements of " Crabs No. 4," the modulus, as
second and
curve.
error,
3*7572)
is
given in
fig.
4 (Plate
as a single group
(i.e.,
16*1914 and
4).
Taking ^ = ygp we have for the fundamental nonic and its first differential
/' (X) = 9X
AX) = X
+ 169-245,583,743 x
+ 24-177,940,535 x
+ 10-054,490,066 x
+ 1-675,748, 344 x
+ 1498'101,518,851 x
+ 299-620,303,770 X
943-393,909,962 x
- 3773-575,639,850 x
864-540 147,350 x
- 2593-620,442,052 x
274'750,163,918 x
- 549-500,327,835 x
34-486,278,563 X
34-486,278,563
1-394,286,418 =
2
0.
* These results were calculated to a higher degree of accuracy than in the case of the Crabs, a result
rendered necessary by the apparent sensitiveness of the roots in this case to a slight change in the value
of the coefficients of the nonic.
Mean
error
is
-7979 a theoretically.
mean
error,
PROF.
Clearly there
is
103
= 2*5868658.
This gave
whence
p2
25-868,658,
Pl
9-669,970.
I found
-Ptf+P* =
were imaginary and no solution involving the difference of two normal components
was possible.
The next stage was to find the negative roots. These were easily demonstrated to
lie between
and 1, and then it was shown that the value of f{\) on ^J changed sign
twice between these values. Thus the nonic was proved, without calculating Sturm's
The two negative roots are
functions, to have only three real roots.
:
Xi=
-154,481,14
-078,262,95.
and
Xa
Carapace of Prawns.
jP2=
Pl
zx
1st
1-544,8114,
26-758,0108,
-057,6086,
y2
-997,856,
=
=
b
=
y =
x
o-i
26-815,6194,
-002,144.
2nd Component.
Component.
cl
(B.)
995,860,
-057,6086;
3-5595,
.111-6142.
ca
6a
= 2-140,
= 26-815,6194,
= 5-7626 v
= imaginary.
7
cr
y2
=
p =
p.z
x
-782,6295,
5-163,5907,
7i=
-147,3614,
y%
-973,0024,
23
%x
= 5-310,9521,
= -026,9976.
104
PROF. K. PEARSOIST
1st
cx
bx
o-j
yx
2nd Component.
Component.
= 971-0564,
-147,3614,
= 3-389,672,
114-28698.
c2
b2
er2
y%
we may add :
Parameters of Normal Curve deduced from
To these
(C.)
solutions
d=
=
=
y =
o
entire
group of observations.
16*191,383,
998,
3-7572,
<r
(D.)
= 26-9436,
= 5-310,9521,
= 8-932,996,
1-203,280.
105*968,04.
tioo
"giants" from
observations.
=
c =
16*14357
=-
-04781),
996,
<r=
3*6051,
y
The curves corresponding
curve are given in figs. 4 and
(6
110-21786.
and
I shall
now proceed
(18.)
The
first
point to be noted
is
The normal curve has a standard-deviation 3*7572, and the mean carapace being
about 43, we have no less than three measurements deviating by more than four
times the standard- deviation from the mean two of them, indeed, differ by nearly
;
six
We
50,000 Prawns, but they are extremely improbable in the measurement of 1000
prawns.
is
in
so improbable that it
We
with a mixed population of Prawns, or possibly there are a few deformed individuals
amid a .normal
There
is
population."'
another point, however, in which the normal curve, based on the total
* I exclude the possibility of any serious error of measurement, having reason to "believe in the great
care with which the determinations were made,
PROF.
K.
the defect
of
carapaces
asymmetrical appearance
about
of the
This defect
45.
curve.
I felt
namely
(see
largely contributes
105
fig. 4),
the
to
the eccentric group of " giants" I could find two components, whose resultant would
fit
the curve of observation as closely as the resultant-curves found for the similar
was peculiarly
whether the method of resolution by aid of the nonic would pay more attention to
the outlying giants or to the less improbable defect of individuals about 45.
imagined that out of the nine possible solutions some might be solutions
As
even
for the
have any meaning are entirely taken up with the very improbable outlying eccentricities
of the observations.
first
observations before the method will be of service in resolving the asymmetry of the
fig.
two giants
is
given.
is
very characteristic,
tion-curve, as far as appearances go, slightly better than the true normal curve.
But the
gives us
first
first
of
which
fits
(giants excluded).
It
component corresponds
ever,
if
an imaginary standard-deviation.
component,
mean
first
mean
This
It has,
how-
first
make no
of 67.
words :
2*14
in
their
resultant-curve (which
must be
observation-curve than
if
it)
is
PROF. K. PEARSON
106
population,
i.e.,
It is
clear
accounts for and covers the dwarf and two giant anomalies
curve, based on the addition of the
two components,
is
an unstable
is
its centre.
is clear,
That
it
first solution,
and a
great advance on the normal-curve C, resulting from the observations as a whole (see
fig.
5).
am
inclined,
homogeneous, but contained between 2 and 3 per cent, of a giant population with a
large standard deviation.
is
may
settle
whether
this
the real state of the case, or whether the anomalies referred to ought to be rejected
when
the outlying parts, which control the nonic at present, are removed.
The
may
all
Table
I.
First.
Second.
Third.
Fourth.
Sixth.
Fifth.
2
3
4
8
27
64
125
216
16
81
32
64
729
4,096
5
6
1-7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
16
25
36
49
64
81
100
121
144
169
196
225
256
289
324
361
400
441
484
529
576
625
676
729
784
841
900
343
512
729
1,000
1,331
],728
2,197
2,744
3,375
4 096
;
4,913
5,832
6,859
8,000
9,261
10,648
12,167
13,824
15,625
17,576
19,683
21,952
24,389
27,000
256
625
1,296
2,401
4,096
6,561
10,000
14,641
20,736
28,561
38,416
50,625
65,536
83,521
104,976
130,321
160,000
194,481
234,256
279,841
331,776
390,625
456,976
531,441
614,656
707,281
810,000
243
1,024
3,125
7,776
16,807
32,768
59,049
100,000
161,051
248,832
371,293
537,824
759,375
1,048,576
1,419,857
1,889,568
2,476,099
3,200,000
4,084,101
5,153,632
6,436,343
7,962,624
9,765,625
11,881,376
14,348,907
17,210,368
20,511,149
24,300,000
15,625
46,656
117,649
262,144
531,441
1,000,000
1,771,561
2,985,984
4,826,809
7,529,536
11,390,625
16,777,216
24,137,569
34,012.224
47,045,881
64,000,000
85,766,121
113,379,904
148,035,889
191,102,976
244,140,625
308,915,776
387,420,489
481,890,304
594,823,321
729,000,000
PROF. K.
D=
IT.
Deviation.
Frequency.
F/P.
o-i
0-2
9950
9802
9560
9231
8825
8353
7827
7262
6670
6065
5467
4868
4286
3753
3246
0-5
0-6
0-7
0-8
0*9
1
1-1
"
1-2
1-3
1-4
1-5
10,
D/S.
F/P.
1-6
1-7
2780
2357
1979
1-8
1-9
1645
1353
2
2-2
0889
2-4
2-6
0340
2-8
0198
0111
0060
0561
3-2
3-4
0031
0015
3-6
3-8
0007
0003
000,004
of breaking
up asymmet-
frequency-curves into normal components has been recognized for a long time
by anthropologists and
R. Livi,
biologists.
vol. 8,
13,
Also by O.
D/S.
0-3
0-4
107
Professor
serious.
'
Firenze,
Ammon
1883,
and
'
Annali di Statistical
Venn have
deaths for each year for 1000 persons born in the same year
curve is also in all probability a compound curve.
the
true mortality-
Since writing the above memoir I have succeeded in resolving this mortality-curve
into components which are not, however,
O.
Ammon,
in the
we
an evolution in
the length-breadth index of the skull of South-Germans since primitive times.
He
does this by comparison of the index as obtained from measurements on skulls from
the Row-Graves and on modern skulls.
frequency-curve
for
up
into
breaking
it
He
to demonstrate
modern South-
108
Germans. In other words, the Row-Graves contain a mixed population, one element of
which corresponds closely to the modern South-German population.
Ammon's
state-
ment, therefore, that an evolution has taken place in this particular skull index appears
The whole problem of the compound nature of skull frequencycurves, both in England and Germany, is a very interesting and difficult one, and I
do not wish at present to anticipate results, which I hope when my investigations
The above may suffice to indicate the range of
are complete to publish as a whole.
to fall to the ground.
problems
to
which a resolution
asymmetrical
of
frequency-curves into
normal
With regard
to the
method adopted
it
in the
memoir
itself, I
am
very conscious
statistics.
may be
am
With regard to
word may be said.
to
me by
have not been able to work out the general problem suggested
George Darwin
Professor
of a frequency-curve,
it
"
may be
dissected
Then
in
mean than
in a size far
it is
obvious that in
same magnitude
in results
two normal
"
is
more
Now
likely to occur in
a low
moment pays
moment
In other words, a frequencyto one far off.
moment
and less attention than a high
curve calculated from low moments fits best near the centre one calculated from
high moments
fits
more attention
tail is
The problem
is
not as probable as an
moment pays
mean
on the other hand, the low moment
to it than a low moment
than the high moment to more probable errors in frequency.
much
observation-curve.
but
if it
;
will prevail
(2r
p. 92,
1) (2r
and
M.
= S (x
2r
3)
we have
.
8a?).
th
for the 2r
5.3.1 o^c,
moment-
45
and
it
(x
aT) (y
6T)
(z
cT) (u
cT)
....
(0)5
represents, besides the plane T, the cubic surface passing through the twelve
straight lines, which are represented in the annexed figure, as well as three other
straight lines which are not represented in the figure.
C
The equations
of the lines
may
~~~
be written as follows
01/ X.
aT
x = aT]
=
=cT
H
,
8)
(10)
u
z
u
I
J-(ll)
= dT
,
2=0^
|>(12)
anci
m
*
d
(13).
""
e
*-
which meets
and (10)
y.= 6T
>(9)
2/
(4)
= dT 1
x = 01
u
= cT
y =
u = dT
=^
2)
6T
=
z
:-
H3
110
PROF. K.
moments
" Crabs No. 4," as calculated from the second, fourth, and sixth
a-2
cr 4
cr 6
Practically, it
would be
theoretical curve.
For
difficult to
moments
= 274,
= 277,
= 2*84.
statistics of this
Thus,
till
is
moments
is
Pear>$6rV:
PhttJrcmsAQ3^A.*jPlate, 1.
West.HewmaiL
sc.
Pexvrson.
Phil.Trans.\W)4>.K.Platz
Wesfc,Hewma,"a.' sc,
i
!
PhiLTrans.l8d4fAJ}lat& 3
Pearson^
V-
~*
--
f -"
'
j"
"
"
t
/i
1
i
vt
-*
f*>
f-
-7
II
It
'
r
I
iV.
1
Sj
SJ
{..
&
^
^
j /
At
NO
te5
'J
^0^
C*
E?
s~
s^iT*
**
S"'
.
^*
^^ ^^
^^ -"
-- ~-
3
.._
""
;t
__
/-' r-
^r.
^ - 7^
-&}-
'
f"
-"
fc/b
^BH-rr
-*"
"^ ^^"
**
=.
7*7
>...
W)
"A
V
v.
1
-^^Si
^*^~
^C
5^*^:. *-*
-.
--:
"-~
*-.
'**
~-
"".
'*"**">.
'
~-^
"
"
*-
^^ *~^
fs
^>
**^*^w.
---
-05
7^.
CO
--
**>*^.
~1
--
>
.-x
i
i
I
.
\f-
-~ -
..__
i
!
_._,
S"
&Hi
'
""
^
'S
'
i
I
i
i
!
j
I
r^t^i"
i
i
_,.
""
i
j
<<
02'
^U
^l Knr,tnArrru
I J}) '&ixy
"""
T
'^T'-
'1
J^i^
i"
>k
i
1
]
j
--
-.
_1
1-
.1
1
.
_L
Wis.t,lf<-jsvTa.%;tt.
sc.
Peccrsorv.
\>
'/
/
k\
/ t
\
i
If
7
f
\
4
//
7
i
<*
Fig.
//
IV.
Length/ of CarccpcLce<
of
Prawns
l!
jI
'
ii1
n 1
p%
R/da R11A
O^tHf,
OII4f'
ft
//
//
/
/
If
*l
i
//
|S
ft
If
.11
it
<h
'
ii
CO
^
1
^
In
^
In
'-1
Oj
//
'Si
i^j-
X'
i
tad..
N
A
>7
^1
^-1
3o
'
$\
//
[
/i
ji
,**1
-^>
-jt&z1
\
J=-^2-
ifis
rig tru"'
Aj xs
T/?
*T
"*
^r
^t
<?,.,-.7.
Phil.
\
\
\
I
\*
\
\
w
i
\\
\v
V
<
\
\\
998
OoservcubLorue
Normal/
Cur\/
/x
x X
X-_x
1 st Component;
5 -<36
2 i4
2'^OympoTheivb
4
\
5.96
38
anxL
42
&jccLLuled/)
StccnxLarcb DeMiaJtixmjs
i 5 * CompoTberub
3.
2^ Comporverub
5 -1626.
NorrwaJb Curve
3-6051
ft
5595
/J
\
1
\V
*
^
1
\H
^
VV
^^
Y'
\^
NV
\V
^=
^s.
r^-Y
Sf*,T 7fl
-?
7-*^
~f
&
An T/1
7
ifZ.;
i
L,
/ o-t~ r-i+V.
'
PhiLTrans. 18 04 .jV.Plate
H
XsSS
1\\
^1
ire-
Ms |\
HIv\W
a "UJ'TU
Adzls
of I/f&OL&ijr&iiverute
16
3377
W- i&im
Origin* cav
Wf.lOOq o^
booby -Z&'vgtiv
4-
Scxxh.
^^
^V^v
SocLLe>: ZtWiits
1V1000 of bo dy
-A
43-00700
~ ler> jgth /.
^^^
West Newman, iith
,
'
Pearsoru
>
/
//
yr**
;'/
'
.7
/
7^
jl
.
/
/
//
.'y
//
h5
/.
/
:7
I
:7
/
Ui /
/
V
/
''
Fig.
Length
2.
V.
17
-
Prawns
of Coorapcuoe of
SdbuddjDTu of Noruuo:
?*=-
//J
-782, 6295.
/.if
JM
if
hi?
',,
:i
I
f
ii
'~
t
I
1
1
'7/
.
hi
Si
ill
ii
'
'/'
rll
"
I fl
I
,
-M
1*
iff
*
jl
i
'
/1
si
Oil
i
.
'
<0
&r
^*-j
i
)
.
'.
...
Hi
t
oi
i?
f:
!
|
'
i
;
I
I
1
i
I
Jr
I
ii
'
41
"f
f\//
'1
~ipw
"
1
i
ii!
1
1
j
i
..
-4
ji
4i
J
i
...
6j
1
'
''//
Fq
t/
w,
/ v/
j.
|
_-
1
i
i
.
...
t
t
i
\
i
l>
1
i
L_
i
1.
1
W.
I
I
i
\
!
j
!
U,
-t
-i
i-
--i.-.
v?
.";
*--f
.,-&>
"*-*
!
Cr
^
t
^rZi
&&sk
(.
.1
J __t
Aip.<s of jMeo
i-
1
!
ii
>-
! U-J
'
1
;
_ x
\-\~-
-j
'
|-
\-
...|
I:....
*\
/"
J<\-
V'\\
//
\
Xr
/
/
s*\
.-
<
vV
X
i
... fc
..-,...,..
X
w
i
\
%
\
t\
re
1
I
iiAJ
!
V
!
No
1
Obs ejrvaiirm.jR
of JridiMukials
998
ft
st
'
911
Comporuenl
2 nr CoTnporueruO
27
V 1
pLesuLtajut/
Normal/ Curv&-*
998
998
...
x *_^
V
'-.-
V.
-,
\
'
l siy
.9//7T7 /i srr*ri\
v. \
RQ 7
Crnrrnmn^i-j
3-
.1
^ Comporusrvb
8-
9330
77
3 ^5 7 2
Normxil; Guurv&
II
.1
'
Oj
i-
ffiH
ii
1
1
oi
\\
JtH
1
!
1
$
^N v
IJ
^ "-TV
Lj_
.
tea
i.
=iMWO
II
_, I
*
1
Ii
i
1
__
i
|
!
i
1
l
"1
'(
<
-7!N
o|
..in.
hScdlZ&:
-v
'i
vs
N
{A
J.
.....j
<
!
i
...
1
i'jl
|\\
1-
c3
zLi)
\\\! V
...q]
-^1
'
\-\\.
!
i,
\~
!
j
!
1
k\ &l
4-*
i
:!
I
\\
fi
V
i
ij
"i
1
1
""'?
*,
i
....
,i
^^
i^^*
^(j
*^fe= rJ^U-
II
1
i
1
i
h od^ -L VTLckth
"n
1
i
j
i
Phil.IrcmJsim^K.PlaJb& 5
dzoals
i
i
i
i
i
1
!
.
1
.1
'
......J......
i
i
'
i
i
i
<
i
j
!
1
1
1
i
1
i
!
^
.
,.,.i
i
i
i
i
'
i
i
'!'
I
i
j
1
i
--^H^
^j-"-**T^~**.
1
1
i
i
4
1
'
1
i
___
_j
>> r-
sj.
i
I
..
Sec JL&
zjn iZ$
. yn 100
07
c h ody
.1
!
"
H^
1
,
1
i
\
l
^v^
-^fe= *
..
^^
.J
i
1
......
*^~
^ _ m
-1 7Uf th
i
'
:l
-*
I
i
i
i
f
i
i
- -
i
!
!
i
.,
il
i
i
1
f -j
i
i
i
I
1
1
i
1
i
!
"
""
I
1
t
I
..
>-*---
'
'
i
1
!
!
1
1
j
_j
\\
I
1
!
i
i
j
,
4-
-!
t r~
ill
III!
i
r
i
i
,
t.
..~
.-,
'!
J.
West/Kfe-'-v-rr: &n, Jj+h.