Professional Documents
Culture Documents
Moments
be the range and
point
S$ r(,,~
3',9. Mo·mcilts. The rth moment--of a varipblc
}' r1 .'
wheIe Z; - X; - i.
In particular
~ - ~ ~, f;(x; - it - ~ ~f;
, - 1
These IeSults, viz., lAc, - 1, 1'1 - 0, and 1'2 - cl. aIe of fundamental im-
portance and should be 'Committed to memory.
We know that if d; - X; - A , then
- ~~f;(d;+A-
, i)', where d;- X;- A
. . . - .!.
N;
1:. f; (d; - 1'1')'
- Ii1 If; [" 'c1 ,,-Ill'1 + 'c2 ,,-2 1'1,2 - 'c3 ,,-3,3
j Qj - Qj Qj 1'1 + ... + (1)'
-Qi 1'1"]
•••(3·18)
- ......' - 'CIJ.'r.-l' 1'1' + 'C21'r-2' 1'1,2 - .•• + (-1)' 1'1" [ On using (3·141)]
In particular, on putting r = 2, 3 and 4 in (3·18), we get
, ,2
1'2 - 1'2 - 1'1
1'3 - 1'3, - 31'2' 1'1, + 21'1,3 •.. (3.19)
1'4 - 1'4' - 4J.L3'1'1' + 61'2' 1'1,2 _ 31'1,4
Convenely,
. . . ' - .!.
N ;
1:. f;(x; - A)' - 1N 1:.f;
;
(x; - i+ i- A)'
1
- N ~, f; (Z; + til')'
Measures of Dispersion, Skewness and Kurtosis
wherex;- X =Z; =
and X A + Ill'
~ #c (~ .rc ~-I
Thus
, _1
Ilr - N ~J; Z, + IZ, III , + rc2Z,~-2'2
J.l.1 + ... + III ,r)
I ~
= Ilr + rClllr_1 Ill: + rC2 Ilr-2 JlI'2 + '" +JlI'r. [From (3·15>:
In particular, putting r =2, 3 and 4 and noting that III =0, we get
Jl2' = Jl2 + 1l1'2
Jl3' =113 + 3Jl2 JlI' + 1l1'3 .. , (3·20)
Jl4' =114 + 4Jl3JlI' + 6112 1l1'2 + 1l1'4
These formulae enable us to find the moments about any point, once the
mean and the moments about mean are known.
3.9.2 Effect of Change of Origin and Scale on Moments.
x-A - - - -
Let II =-h- , so that x =A + hu, x =A + hu alld x - x =h(1I - u)
I -
=hr IV J;./; (u; -
I
uY
Thus the rth moment of the variable x about mean is h r times the rth moment
of the variable II about its mean.
3·9·3. Sheppard's Corrections for Moments. In case of grouped
frequency distribution, while .calculating moments we assume that the
frequencies are concentrated at the middle point of the crass intervals. If the
distribution is symmetrj~al or slightly symmetrical and the class interval.s are
not greater than one-twentieth of the range, this assumption is very nearly true.
But since the assumption is not in general true, some error, called the 'grouping
error', creeps into the calculation of the moments. W.F. Sheppard proved that if
(i) the frequency distribution is continuous, and
(U) the frequency tapers off to zero in both directions,
the effect due to grouping at the mid-point of the intervals can be corrected by
the following fonnulae, known as Sheppard's corrections:
h 2
Jl2 (corrected) =Jl2 - 12 ... Q·21)
Jl3 (corrected) =113
a I -- H:!
(J -
- 0, (X - 112 - 1
2 - (J2 - ,
(X _l!2.
3 ~cP -- -v-fa -
PI -
y 11 (X _ 114 - f.I.
4 - (J4 - 1:'2
II( )'
rr
=1- i: 1', X .<r)
N;=I'"
... (3·23)
1/
t "J;./;,
J ' •
11(2)' t
= "J;./;xP)
, = "J;./;x;
, (x; - t 1)
Measures ofDispersion~ Skewness and Kurtosis 3·25
J.1(3)' = ~ "i;./; xl:') = 1
N "i;.f~i (Xi'" ) )(Xi - 2)
r r
=~"i;./;Xi(X?-6x?+
r
1]xi- 6)
I I I I
= IV "ffi Xi4 - 6. IV I./;x? + I J. IV I./; x? -6. tv I./; XI
= J.1/ -
6J.1/ + IIJ.1i' ~ 6J.1.'
Conversely. we will get
Ill' =J.1(l)'
J.12' =J.1(2)' + J.1(1)'
J.1/ = 11(3)' + 311(2)' + 11(1 >' '" (3·25)
J.14' =11(4)' + 611(3)~ + 711(2){ + 11(1,'
3·12. Absolute Moments. Ror the frequency distribution XI if; i = I. 2•...
II. the 11h absolute moment of the variable abo¥t ~he origin is given by
I"
N r= •
r
- .I. /; Xi N = I./; 1 I. ... (3· 26)
where 1x[ 1represents the absolute or modulus value of x[ .
The 11h absolute moment of the variable about the )1lean x.is g,v~n by
I " -
-N'I.
r= •
/; 1Xi - x l · · · (3·26a)
r
Example 3·8. The first four momellfs of a distribution about the vallie 4 of
the I'llriab/e are - J·5, 17, - 30 and, 108. Find ,the moments t'1/io/lf liiellll, PI lind
f3!.
Find (//so the moments nbollf (i) the origin. and (ii) the poillf'x =2.
Solution. In the usual notations. we are 'given i\ '= 4 and
J.11' = -1'5. J.f2' = '17. J.1/:: - 30 and 1I4' ='108.
Moments about mean: J.11 =0
J.12 = J.1{ - J.11'2 = 17 - (-1·5)2 = 17 - 2·25 = ]4·75
113 =113' - 3112' J.1;' + 2J.1{'
=.- 30 - 3 x (17) x (-1 ·5) +.2 (-1·5)3
=.- 30 + 76·5 - 6·75 = 39·75
Fundamentals of Mathematical Statistics
~2 =Il~ =(142.3125)=0.6541
112 (14·75)2
~I(= NIl;
1- '1 Xik 1 ... (1)
Let u and I' be arbitrary real numbers, ,then the 'expressfon'
II 2
I. Ii [u' XP:-I)J2, + \ll.xP:+I)J2,] "is nc;m-g~gati~e,
i. I
II 2
~ I. /; [U I Xil(k-I)/2 + \II Xi l(k+I)J2] ~ 0
i- 1
~ 112 L/;lx,lA>-l + \1 2 Lli lxilk+l + 2u\I "'i:.li lxil k ~,O
Dividing throughout by N and using reiatiOll (1), we get
,,2(3"-1 + v2 (3"+1 + 2uv~" Ot 0, i.e., '?~I(-I + 2uvl3" + v:! ((+1 Ot 0 ... (2)
We know that. the condition for the' expression (/ .\'2 + 2/lX.'':'+ b i to be
non - negative for all values of x llOd y is that
la hi OtO
I h b I
Using this result, we get from (2)'
I~: - 1 !~: +1 I~ 0
pomt ·'c?
.
and, =0 otherwise. i.e.. the whole mass of the variable is concentrated at a singie
•
Since P (X =e) = I. Var (X) =O~
Thus adegenerale r.v. X is characterised by Var (X) O. =
M.g,(. of degenerate r.v. is given by
=
Mx (t) =E (e IX ) =e'c P(X =c) eCI ... (7·Jb)
7'2. Binomial Distribution. Binomial distribution wa~discQvered by James
Bernoulli (1654-1705) in the year 1700 and was first published posthumously in
1713. eight years after his death). I,.et a randqm experiment be performed
repeatedly and let the occurrence of an event in a.trial be called a success and its
non-occurrence a failure. Consider a set of Il independent Bernoullian trials (Il
7·1, Fundamentals or Mathematical StatiStics
being finite), in "'1hich the probability 'p' of success in any trial is constant for each
trial. Then q = 1 - p, is the probability of failure in,any trial.
Ti.c probability of x successes and consequently (n -x) failures in n inde_
pendent trials, in a specified order (say) SSFSFFFS .. .FSF (where S represents
success and F failure) is given by the compound probability theorem-by the
expression:
P (~SFSFFFS .. .FSF) = P(S)P(S)P(F)P(S)PH1P(F)P(F)P(S) x
••• x P(F)P(S)P(F)
s p . p .~ .. p . q . q . q . p •.. q . p'. q .
l
. ='p,p ....p
1
- Butx successes in n trials can oC('\lrjn ( ~}ways and the probability for each
of these ways is PC'cf'-x.. Hence the probability of'x successes in n trials in any
order whatsoever is given by th~ a"d,ition theor~m of'prqba hil JlY by the ~xpression:
x
Th ·l· d· 'b .
. epro bab Ilty- Istn utIOno t6enum ro
'(f;)PX
cf'-be f
successes~so.o
' • b . d· . I
tame 1SC3 led'
the Binomial probability distribution, fOI\ the ,obvious riason that the' probabilities.
of 0, 1~ 2, ... , n successes, viz., ,
: (n),/I'-
' ,I «
,/I n) «
1 p, '( 2 ,/1.- 2 . terms 0~ f t,be b·mo·
p 2, ..., ,P" ,are t. he successive
·(~ r( 1~
=
120' + 45 + 10 + 1
) t (
176
=--
~'J ('~o.) + ( :~ )}
t
1024' 1024'
. . , , ' ~I", .
Exa~ple' 7·2. A and B playa gllme in wli;ch tlleir cl~a{fces ofwinning qre
i" the ratio 3 : 2. Fin~ A's chance l!f winning at least ,three gqmes., outl?[ t!ie five.
8ame~play¢d. ' [8~rd~an Univ. p.S~. (Hons,),199~]
, Solution~ Let p be the probability that 'A • w1ns the game. Thc:P. ~e are
glvenf = 3is' :=T' q = I-p = 2/5. ,
"Hence. by 'binomia1 probability law. the probability Jhat out of 5 gant~'
played.A wins 'rl.gilmes is given by :
7·4 FundamentaJ.s or'Mathematical Statistics'
The required probability that 'A' wins at least three games is givenby :
P(Xi1:3)= ~(
,-3.
5) 3' .~5-'
r 5
Example 7~3. If m things are distributed among 'a' men and 'b' W{>men,
show that the probability that the number pFihings received by men is odd, is
.! I~ (b + a)m :.....(b _ a)m]'
2 l
(.b + a)m
(Nagpur Univ B.Sc., 1989, '93)
Solution. p = Probability that a thing is received by man ... ~b' then
a +
q = 1 - P .. 1 - ---.!!.-b
a +
= ~b'
a +
is the probability tbat a thing is received by
woman.
The probability that out of m things exactly x are received by men and the
rest by women, .il1 given. by
p(x) .. mCxIfq"'-x; x ... 0,1,2, ... ,m
T~e probability P that tlie num~r of things received'by Illen is odd is given
by
P • P(' I) + P(3) + P(5) + ... - "'cI' q,"-I 'p + '" C3' q",-3 'p 3 + "'c5' q.. -5 'p5 + '"
Now
+ I'q... -1 'p.+ "'c2'q",-2 P2 + "'c3'q",-3 "p3 + "'c4'q",-4 'p4 + .. ,
(q+p)'" -q"''''c
and
( q-p ) '" -qIII -''"c"q",..,1 'p+'"'c2'q,"-2 'p2 - "'C,
3,'q,"-3 'p.3 +'"'c4'q111-4 'p4 - .. ,
., (q + pT -(q-pt =.2 [mCI . q"'-l.,p + mC3' q"'-.3·l + ... J= 2P
b-a
But q + P = 1 and q - p = -
,., ,..-a
=>
=> 0'01
1 - {(
OP: ~
~ ) + ( ~ )} ( ~
=>
r OP:
2ft
'2!' OP: 100 '+- 100 n-
By trial method, we find that the inequaJity,(*) is satisfied by n - 11:Heoce
the minimum number of bombs needed to destroy the target cOmpletely is 11.
Fundamenta~ of Mathematical StaUS•.lcs
.' 'T,he probability that "lfjust two machines need adjustment, they are oftbe
same type" is the same as the probability that "either just 2 old and no new or just
.2. I).ew and no old machin~s need adjustment".
. :. Required probability = 0·016 + 0-028 = 0·044
. . 7; 2· 1 l\'1oments. The first four moments about origin of binomial dis-
, '~~!i~'t.!ion ~re obtained as follows:
1'2' .. E(r) s i:
x- 0
x2
~
( n ) pX
X
c/' - x
=x ::)x (x -
If 11 (n - 1) In - 2 \. .
1) + xl x (x _ 1) . x _ 2 pX q'7 x l J
c n (n - 1) p2 [x ~f (; =~) pX- 2 q"-x] + np
=n(n-l)(n-2)p3 i (n-3)px-3c/'-X
x - 3
x - 3
+ 3n (n - 1)/ r
x-2
(n - 22) px-2 q"!.ox + liP-
x- ,
= n(n - l)(n - 2)i (q + p),,-3 + 3n(n - 1)/ (q + pt-:. 2 + "I?
= ~ (n - l)(n - 2) l + 3n (n - 1) p2 + np
Similarly
X4 = x (x-l)'(x -2)(x-3) + 6x (x -l)(x..!. 2) + 7x (x-I) +x
, Leti. Ar(x-l){x -2)(x-3)+Bx(r-l)~-2)+Cx(x-l)+x
, By giving to x the values ~1, 2 and 3 respe~tively, we find the values of
arbitrary constants A, Band C. Therefore, ,
E (! "'-
n
2
p) ... I!9.. Cov
n' n'
(!
!!-=..
n)
!\.. _I!9.n
(Delhi Univ. B.Se., 1989)