KEMBAR78
Moment Generating Function | PDF
Scribd
Upload
151 views20 pages

Moment Generating Function

Uploaded by

Aditi Singh Bhati
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
151 views20 pages

Moment Generating Function

Uploaded by

Aditi Singh Bhati
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 20
, Acbastote convengeare,— un eo non Bee A Bijan Sa a Fevieys w Sp : FUNDAMENTALS OF MATHEMATICAL gran, © 29 Ce. ane L Care? OL geo (fer = L (fia 77. Gea OF LARGE NUMBERS 78. BOREL-CANTELLI LEMMA _ 7.9. PROBABILITY GENERATING FUNCTION 7-9-1. Convolutions CHAPTER CONCEPTS QUIZ / DISCUSSION & REVIEW QUESTIONS / ASSORTED REVIEW PROBLEMS FOR SELF-ASSESSMENT “MOMENT GENERATING FUNCTION The moment generating function (m.¢.f.) of a ha bility function f (x) is given by : | ef (x) dx, (for continuous probability distribution) Le= F(a), (for discrete probability distribution) eau =E(e%) = | (7A) | the integration or summation being extended to the entire range of x, ! being the real parameter and it is being assumed that the right-hand side of (7-1) is absolutely convergent for some positive number |i such that — ht < ¢ [w Fa we | (73) In general, the moment generating function X about the point X =a is defined as: oe A 8 My ees Or femnider ini He ASG Os Hastagente Gy Ga a pee 7 tenet p BA hiv trgent or ESA 5—Snus con df, X57 a, COB.LARAE NUMBERS Mx(t) (about x = a) SE (eck i es Se MAE ay v 7 | , ai aay, y " 1 R ri ay) ‘ = ED ie TAP ie eta | ¢ Te NN Pipa (74) | where bi’ = E {(X—a), is the rth Moment about the point x 4 eI 7-1-1, Some Limitations Of Mo na . Y ‘ nt Generating e eens Peon Bulfea from some drawbacks which ta coc ans ‘Statistics, a give 4 na Some of the deficiencies Of m.g.fi's with illustrative examples, | 1,A random varia eX may have no moments although its M8 exists. For example, jet Useonsider a discrete ry, wine Probability functions a ah Bconsider a discre ? 1 es: F@PHX=12,,, 0, otherwise Ew =D, Se Ree oft au S(t Mee FQ) 2) fe) way hh. eae sat ieee Since, x x 15a divergent series, E (x) does not exist and consequently no pee > —— tera toment of X exists. However, the m, '8:f- Of Xis given by : i erica Fer erty =e) ~log (1-2) +142 10g (12), bz) <1 =1+(1 =1) log (1-2, lz 1 <1 Alt(et Also Mx(t) = 1 for t = 0. Hence My(t) exists fort <0. ~1) log (1~e'), #<0 [From] fs Iz eteteler =1<0) Ce=2 =F uy ee x! } EQ)= 2 ay pKa2y=01 L S01. exp2)= ep 2-1) ne Oc in} Mea), 9 Hence all the moments of X exist. The m.g.f. of X, if it exists, is given ch Se = BB) wi (t.2957 Mx) = % exp(t.29(G)=e1 D ett 297 ‘By D’ Alembert’s ratio test, the seri oe the series on the eee pe f > 0. Hence Mx(t) cannot be Woe COnveRgeS fog | turin’s expansion and consequently it does not fated catiraty a | 3-4 ro. X can have all or some moments WE, até moments, “4 ha te -For example, let X be a r.v. with Probability function ; a P(X =+2%) -[Brs8 ea 0, otherwise Since the distribution is symmetric about the line X = aboutorigin vanish ie, EX") =0 yu, ano,” OMENS oF odd oa . t (ee oy 1 e = xl =e &XP(2")=exp (2-1) > i e x)2r( 1 E (x?) = (£2%) ( 7 Thus all the moments of X exist. The m. &-f. of X, if it exists, is given by Mi) = & (0 ret) gh] wet cay x=0 a which converges only for t = 0. As an illustration of a continuous probability distribution, consider Paret 2 0 distribution with p.df. p(x) =o x20, 0>1. EQ) = 004 xt 78 “Idx = 608 a 16) x , Which is finite iff r-8<0 + @>y r-O"", and then Ex”) = 048[0- #2] 28 gs, r-0 6 However, the mg,f. is given by : Swe Mx (ft) =8 ao zist 4x, which does not exist, since e* dominates x°+! and a (e% / x° +1) — as x > ~ and hence the integral is not convergent. For more illustrations see Student's t-distribution and Snedecor’s F-distribution, for which m.gf-’s do not exist, though moments of all orders exist. [c.f. Chapter 15]. Remark. The reason that m.g.f. is a poor tool in comparison with characteristic function (cf. § 7-3) is that the domain of the dummy parameter ‘t’ of the m.g.f, depends on the distribution of the r-v. under consideration, while characteristic function exists for all real t, (- H/o My (t/a) vo (78) fh es Remark ( E(Z) =(454) -Ler-w=! 6)-we2 w-w-o and x Ve) =v (254) =2 vex-w =2vc-2e «, E (Z) =O and V (Z) = 1, i.e., mean and variance of a standard variate are 0 and 1 respectively. 7-1-3, Uniqueness Theorem of Moment erating Function. The moment generating function of a distribution, if it exists, uniquely determines the distribution. This implies that |corresponding to a given probability distribution, there is only one m.g,f. (Provided it éxists) orresponding to a given m.g,f,, there is only one probability istributior Hence My (f) = My (t) = Xand Y are discussion, see Uniqueness Theorem of Characteristic Functions.] 7-2. CUMULANTS Cumulants generating function K (t) is defined as : Kx (#) = log, Mx (1), Provided the right-hand side can be expanded as a convergent series in powers of t. Thus Kx) Bayt age tt Koy + = log My (D) x he ai = log (1 + py't +h or Fi 3p te +My ppt 2] «+» (7-9a) where x, = coefficient of 4+ in Kx(t) is called the rth cumulant, Hence lentically distributed. [For detailed tt (Be pase a(uresuy fn d endart ) b(t wd a thy sy is i ici i f ‘t’ on both sid C the coefficients of like powers 0 a benltaahip Between the moments and cumulants. Hence, we have Bet the ie ' @ Pe 2 ki =yy'=Mean, 2-82 - Dr = ky t’—hy? = be Bye WT PATER Beds Sad +29 =, (1 (ug? , 2mi'es) , 2 Saat _ wa Also Ma 1 (me 2B) «5 Sue = Ky = Hy 3 py — 4 py’ Hy’ + 12 pr? fy’ - 644 = (Ug —4pig py’ + 6pl2’ Hy? — 3pts’4) - 3 (42? = 2p2’ Hy? + 14) = Hg —3 (Ho! — x2)? = Ha — 32? = Ba = Ske? (snes . Py = Ky + 3 Ky? Hence we v= (7-9) Remarks 1. Thes® memory. 2. If we differentiate both sides of (7:9a) w.r. tot ‘r’ times and then put t = 0, we get : © of fundamental importance and should be committed to eas (7-9¢) 7-2.1. PROPERTIES OF CUMULANTS foperty 1. Additive Property of Cumulants. The rth cumulant of the sum of the independent random variables is equal to the sum of the rth cumulants of the individual variables. Symbolically, Ki, (X1 + Xp + ov. + Xn) = Ky (Ky) + Ky (Xp) + 0 + Ky (Xy), + (7110) where X;;i=1, 2, ..., n are independent random variables. Proof. Since X;’s are independent, My, 4.x, 4... x, (t)= Mx, (t) Mx, (#) « Mx,() Taking logarithm of each side, Kx, 4 x, 4.x, (t) = Ky,(l) + Ky,(l) + + Kx) Differentiating both sides w.r.to. t ‘r’ times and putting t = 0, we get [sek xy + +20], -[are0] + [Sx .0 |_t + [ex | Ke (Xa + Xo + oe + Xp) = K(X) + K(XQ) +... + K(X), which establishes the result. b= 0 ID LAW OF LARGE NUMBERS a4 * Effect of Change of Origin and Scale on Cumulants. If we tae ' uy =X then Mu (0) = exp (alli) My (i) Ku (t) = log Mult) =~ 4 + Kx(t/h) i ‘i yewh = +e ee +e we ¥; (I/h) + Ky ty, wath Kp “hy - i hax? and x, are the rth cumulants of U and X respectively. : Comparing coefficients, we get [ry’= A and k/=feir=2)3,.. (740 ‘| t the first tumulant, all cumulants are independent of | igin. But the cumulants are not invariant of change of se the rth | Uis (1/h’) times the rth cumulant of the distribution of X. pian) =4-1P iT = 1, 2,3, Find the mgs. of X and hence its mean and variance. ote ae solution. Mx(") =E(e%)=% er g-tp=t E (gl = Fae x (qety -1 = pe! (1+ ge! + (get) +. ( 3) Ifdash (’) denotes the differentiation w.r. to f, then we have a pet five fue Gene) Mx) = Gagan’ Mx'(t) = pe ap 11’ (about origin) = My’ (0) = aor 2 ; , in), = My" (0) = LOO wy (about origin) = Mx” (0) = G@—ar = T3t P ’ fF earenal tule ea Hence, megyr= jy’ (about origin) =} and variance = y= Hy’ —bi?= poy = J 5 7-2. The probability density function of the random variable X follows the 1 lx-6 | probability law : p (x)= Jexn(- SS) -20 1x0! =0-xVxe (-~,@). Similarly, | x-@ 1 =x-OVxE @-)) ep {x(t 4} dre al exp {-=(4-1)} dx : FUNDAMENTALS OF MATHEMATICAL Bran log a eadaacas) eo 20 (v i) tt of Oa ol (1 = O22) “Fore ago “108 3042 2( 1014 Be +) OOM, DAM eT Soe ~ Var (X) Ha! = by? = 80? = 07 = 2. 0°. le 73. Ifthe moments oe ‘variate X are defined by E(X") = 0-6 yr = 1, 23. P(Xm0) #04, P(X=1)=06, P(X22)=0. | | EQ) = Coot, ae (=O, ) p= Coelt of pin Mx (t) =3 ‘Solution, The mgfof variate X is: wet Mx() = zi rte + be H (06)= BO Oe S fr = 044060 a rnd 2 i But Mx ()=E(e%)= Le P(X =x) 1 xed =P(X=0)+e'-P(X=1) + ie et. P(X =x) (*) From (*) and (*), we get: P(X=0)=04; Pee 1) =06;P (X22)= Example 7-4. Find the moment generating function of the random variable rohose moments are: ‘ Hy s (r+ 1) 12", Solution. The m.g,f. is given by : Mx() = + wed ff (r+ tar= x (r+ 1) Qn reo =14+2 (21) +3 (2H2 +4 (21) + Aliter. The RH. is an arithmetic-geomet Let S$ =142r+ 324409 + Then TS = rt 2438+... (l=) S=14rt ete 1-28). ies with ratio r = (2). y= (1-242 TY Remark. This is mg, n=2 (cf Chapter 15} Example 7-5. If i, is the rth moment about origin, prove that f of Chi-square (x*) distribution with parameter (degrees of freedom) the jth cumulant, : ‘ r-1 we = E74) es ttere Solution, Differentiating both sides of (7-92) in page hs oa w.r.to. t, we get fate 2 Hy! + py! t+ py’ a + by PB rt =a! Kp EP aT Hath, Gott w= re ) 3 14+ ty’ t+ py" wt at Uy opt a “pplica tm Auch exp. Uhoke qoe QELEMSAES GH ents ME Lavatt ice gpeciAL DISCRETE PROBABILITY DISTRIBUTIONS ye 8:3 experiment. The present study wi 7 otncton of the form y= p(x) to uae enable us to fit a mathematical model or data. is ter is devote , a ee the study of univariate (except for the multinomial) ¢ Pern Wulunonie! and Paved disson, Negative Binomial, Geometric, Dred distribution function, mathematical exrect distributions. We have already distributions. 4.2, DISCRETE UNIFORM DISTRIBUTION Definition. A 1.0. X is said to have a disc rete uniform distributi i ite pm, is expressed as follows uniform distribution over the range [1, n] if ‘forx= 1) Dy P(X =x) | 0, otherwise vs (8D) Here n is known as the parameter of the distribution and lies in the set of all positive integers. Equation (8:1) is also called a discrete rectan ign. Such distributions can be conceived in practice if under the given experimental conditions, the different values of the random variable become equally likely. Thus for a die experiment, and for an experiment with a deck of cards such distribution is appropriate. ez 8-2-1, Moments. E(X) = Jar. Distt) Ext Sire (n+ 1) (2n +1) fl Si 6 n 2 + AV) = £00) ~[EQy= += et) n(l—e) “The mgfofXis: Mit) = E(e%)=3 Nas berNouL DISTRIBUTION Definition. A r.v. X iss uli distribution with parameter p if its p.nuf. is} given by : | [ee (1-p), forx=0,1 | | pa P(Xeoa) | 0 athentiss os = The parameter p satisfies 0

7) =p(7) + p(8) + p(9) + p(10) 10 K 1 10) , (10) , (10) , (10) _ 120445 +10+1_ 176 =(3) {(:9)+(#)+(®) +(8)}- 1024 a” Example 8-2, A and B play a game in which their chances of winning are in the ratio 3 : 2; Find A's charice of winning at least three games out of the five games played. Solution. Let p be the probability that ‘A’ wins the game. Then we are given : n=5,p=3 | __ Hence, by binomial probability law, the probability that out of 5 games played, A Wins ‘x’ games is given by : Sealy le | Zor Biornet GANT Meo F Nain, FUNDAMENTALS OF MATHEMATICAL STATISTigg ® sien pegby =ne=(3)(3) (O* ra ‘The requird probability that ‘A’ wins at Jeast three games is given by : Sa S3) = 3} ae =) 2+(§). ax241sex1| 27x (40+30+9) _ g.6g “A 315 5 ; i i distinguish between a c Example 8-3. A coffee connoisseur claims that he car 4 a eup of instant ee a Eee sien coffee 75% of the time. It is agreed that his claim will be accepted if he correctly identifies at least 5 of the 6 cups. Find his chances of having the claim (@ accepted, (ii) rejected, when he does have the ability he claims. Solution. If p denotes the probability of a correct distinction between a cup of instant coffee and a cup of percolator coffee, then we are given : 1 at p= -2 = 4q=1-p=z, and 1=6 If the random variable X denotes the number of correct distinctions, then by the Binomial probability law, the probability of x correct identifications out of 6 cups is Ey x 6-x P(X=x) =p@=($)(3) (4) PSO lA Zest ( The probability of the claim being accepted is ! rasy=ns)+t6)=( $)(3) (4) +(8) (3). = Hse + Bs = 058 e probability of the claim being rejected is : P(X <4)=1-P (X25) =1-0:534 = 0-466. xample 8-4. A multiple-choice test consists of 8 questions with 3 answers to each tion (of which only one is correct). A student answers each question by rolling a balanced die and checking the first answer if he gets 1 or 2, the second answer if he gets 3 or 4 and the third answer if he gets 5 or 6. To get a distinction, the student must secure at least 75% correct answers. If there is no negative marking, what is the probability that the student secures a distinction ? Solution. Since there are three answers to each question, out of which only one is correct, the probability of getting an answer to a question correctly is given by : p=1, sothat g=1-p=3 By Binomial probability law, the probability of getting r correct answers in a 8-question test is given by : P(X= x) = p(x) =(8) Gy. ey F850, 1p2p5ac 8 Hence, the required probability of securing a distinction (i.e., of getting correct answers to at least 6 out of the 8 questions) is given by : ro -r0er6) -(3) 0) +0) G0)" ETE PROBABILITY DISTRIBUTIONS ea 89 He 8-9. In a binomial distribution of 1 and 2 successes are 0-4096 and (). distribution. Solution. Let X ~ B(n, p). In usual notations, we are given : n=5, p(1)=04096 and’ (2) =0.2048, According to Binomial probability law : POX = 3) = pe) =(8) pr pps, 2-0, 1,2,...,5 Now p(1) =(3) p(a-p)'=0-4096 zat) and p(2) -{ 3) p21 -py= 02088... (*) Dividing (*) by (*), we get () POP) _ 0.2056 () sags 02088 2) B (1 -p) Example 8-10. With the usual notations, consisting of 5 independent trials, probabilities 2048 respectively, Find the parameter ‘p’ of the 5(- => Se =02. ae > p= find p for a binomial variate X , if p=6and 9P(X=4)=P(X =2). Solution. For the binomial random variable X with parameters n = 6 and p, the probability function is : P(K=1)=(8) prq’";7=0,1,2,..46 Weare given: 9P(X=4)=P(X=2) 3 9x ($) pig? = (8) p44 > Sp? = @P oo =e py Laer => 8p? + 2p-1=0 => =-7t ce probability cannot be negative, p =~ }is rejected, Hence p =}. 8-4-1. Moments of Binomial Distribution. The first four moments about origin of binomial distribution are obtained as follows : x (a) prt gh = np (q +p) =np [-@)-2@) = 258 (123), andsoon Thus the mean of the binomial distribution is np. wy’ =E(X) = D x(") pr =np 0 ws = B00) = Bo x2(") pegrs = D (x(x-1) +x} 2) (223) a= = x(x=1) | cri { e (2:3) p20} +np ae en. FUNDAMENTALS OF MATHEMATICAL STATISTigg n(n 1) p2 (7+ pyre rnp = (1) PP by =E(X)= . 8 play= 3 ae 2) +30 6-14 81(2) peg 0 en(i-1)(0-2)P x een? 4+ 3n (n -1) p? 2 (24) Pg + np =n (n=1) (02)? (g + pyr? + 3m (2-1) (+ PY? + Op =n(n-1)(n-2) p? + 3n (n—1) p? +P Similarl} | a =x (x-1) (x2) (x-3) + 6x (x -1) (1-2) 47x +x Let xt = Ax(x -1)(x-2) (x-3) + Bx (x-1) (x-2) + Cx“ +x (By giving to x the values 1, 2 and 3, we find the values of arbitrary constants A,B and) Ae HERS = z aA(1) pa eee. 6n(n-1) (1-2) 2 + 7n (n-1) PP + mp =n (n=1) (2) (n—3) ph + 6n(n—1) ( {On simplification) | Central Moments of Binomial Distribution « A hy = pg = 2 = rp np? + np — np? = np (1p) = npg yA Hs = Hs’ — 3dr’ by’ + 2p? = {n (1-1) (1-2) p? + 3n (n—1) p? + np} -3 {n (0-1) p? + np} np +2 (mp) = np (—3np? + 3np + 2p? -3p + 1 ~3npq) . j = np {3np (1—p) + 2p?- 3p + 1 - 3npq} = np (2p2=3p +1) = np (2p? 2p + q) = npq (1 - 2p) = npq {q + p—2p} = pq (q-P) Mtg = Ha! — Apts! by’ + Olly’ py? - 3px = npq (1+3(n-2) pq} [On simplification] Hence 2 5 = 2 eee fap _ (129 e “a wep P "pq Pq tL = Ye _ mpg (1+3(n =2)pq) _1+3(n-2) pq_ 3 , 1-679 (86) : Bae PE me Be (86) | Begg a2 1, 9g 5-1 =a aan ih A ieee ae tabs a Roefnarks 1. If X ~ B(n, p), then mean = np and variance = npq “Hs =npq(q—p) and y= npg [1 +3(n-2) pq] (6-64) 120 Variance = npq P=3- Substituting in (*), we obtain n= P(X21)=1-p (X=0)=1-qr=1-( ) (edample 8-13. If X ~ B (n ,p), show that : 2 E(X-p) = @, Cov (X, wax n nT Solution. Since X ~ B (n, p),E(X)=np and Var (X)= npq ye 1 e(4) = FO) =p; Var (2) = <9. Var (x) =P aw) z x (i. E( E( ) all [From (*)] x n x < at PNA - ‘ E(%) =~ Var (7) =-t. 8.42, Recurrence Relation for the moments of Binomial Distribution (Renovsky Formula). By def., jt, = E (X - E (X)" = & (x —npy( 1p a 0 (i) Cov( Differentiating w.r. to p, we get he = (@) [qr (= mp) ps gt* + = npy {xp gh = (1-2) peg] 0 Beas, (2) (x— np) pt q+ x CG) (e=npy pears (% x=0 = =—nr X (x—np)" p(x) + x ese “a x0 a odd = (2m + 1), say. Then FDP =Qm+2)xF< 41 (integer)="=14 1 n+l iz + 1) p is an integer, the distr - , ib iL (n+1)-1 anus i ‘ution is bimodal, the two modes being Mot i ment Generating Function of Binomial Distribution. Let X ~ B (1, p), Mx (t)= E(e%) = E en( nN pegs = (") (pak ge Sgeeene Meat m.gf. about Mean of Binomial Ditibeion : Elem) = e-tnp . E(e%) = et : r em} (e%) =e - My(t) =e"? (q + pel)" = (ger + petty" fiat Z p =[a{1- pt+ BP ee, ef. | eS ee Seay 31 +) + 57 Pa (9+ p) +4 pg (P- PY +E pa P+ P+. y al [a+ {Frat dpm p)+ fap (t- 3pq) +.. j] =[1+(7 ){ Fep+Sea- p)+ rg a- 3pq) + | (3) (en pa+tepq G-p+ ole el 2 3 Now hp = Coefficient of = mp4, Ht = Coefficient of ; {ly = Coefficient off; = npg (1 3pq) +3n (1-1) PPP = Sn? pq? + npq (1 6p9)- Example 8-17. X is binomially distributed with parameters n and p. What is distribution of Y =n -— X ? __ Solution. X ~ B (n, p), repre: with constant probability p of success for each trial. sents the number of successes in 1: independent trials ATICAL Sta in n independent eY=(n-X)~B(n,q) ‘als My ()= E (e) = (q+ Pe)" = ott, (C18) = oH MH) é Cae Gage) JQ) 9 + Solution. Since My (!) -( 2,14) = (9 + pet), by uniqueness theorem ofmee . = = 1 X~B(n=9,p=3), Hence E (X)= px = "p= 3: Ox! = PAH 9 xT x29 pt2o0 =342x¥2=342x1-4=(0-2,58) 2 P(u-20, Generalisation. If X;, (i = 1,2, ..., k) are independent binomial variates with k k parameters (ry, p), (i = 1, 2,...R), then their sum ¥. X,~B ( sy nua) ia ia The proofs left as an exercise to the reader, Example 8-19. If the independent random variables X, Y are binomially distributed, respectively with n = 3, p=3" and n=5, p=1. write down the probability that X + Y 21 ROBABILITY DISTRIBUTIONS oar aregiven: X~B(3, 4) and ¥~B (5,1), | i lieerysttiied eas ee rien ibe eddies property of binomial datrouton, we ‘ X+Y~ B@+5,}), ie, cr. B64) a PX+Yena( 8 yay ue?) Hence P(X+Y¥21)=1-P(X+¥<1)01-P(X4 Ya O)e1- ( 3)" prom (9) __ 48, Chaocttte Function of nome Dbton. = IX) x(t) = Ee#%) E ap z x (2 2p grea E (2) wera =(q + pet (8-13) | \ Spe Cumuiants of the Binomial Distribution, Cumulant generating function is given by: Kx(t) = log Mx(t) = log (q + pe'y" = n log (q + pe’) =nlogg+p(1+t+ Betis +h. ) anlog{i+p(t+f+hete.. ) 2 =n[p(t+ BeBe te.. . ‘4 Ble fe) Affe Sen] Mean = k, = meee rot in kel np Hp = Ky = Coefficient of in Kx(t) = (p-p) = np (1-p) = npq The coefficient of #8 in Kx(t) an[2-B od +2] = Fa- 3p + 2p?) k= Coefficient of Fi in Kx(t) = np (1 - 3p + 2p2) | = np (1-p) (1- ws = npq (1-p-p) = npq (q-p) ee Hg = Ks = "pq (q-p) The coefficient of #4 in Ky(t) 23 ta By Ka wn[ fe (Gro) Bah Eero apo K_ = Coefficient of 3 in Ky(t) = np (1~p) (1 6p + 6p") = npq [1 ~ 6p (1~p)] = npq (1 - 69) FUNDAMENTALS OF MATHEMATiCg, ey an isp, Ss By = BE = pg (1 — 6p) + 3H? P? = HP CA ~ 679 + By p99 ve i = npq [1+ 3pq (n= 2)]. 8.4.10. Recurrence Relation for Cumulants of Binomial Distibution x, = [frlog My] a [fs log (q + re] im "BY dg de, fat ad »| =n[ gp C142) * a a n[ ap 108 (9 + Pe!) pete 7s, a fal ie Kr =n] tog (a+ pel] ok? | qos 4 + pel} oe of (8) = nla apa), —"aar (sane), 0 Hence dk, (Sue [fr dol | knd [z Lt pel i aa geepely | e® _* [s | 3 0 AO 0 where G(E) is the probability generating function of X. Find it when X ~ B(n, p), anda = 1. 1 1 Solution. RHS. = | » ee GU ae -| 1 E0%) at ° =f, {e( Er) = Bhp ea Ee. sarelga) 0 @+a) 1" IfX~B(n,p) then G(t)=D p(x) =(q + pry" x0 Hence taking a= 1 in (*) and using (**), we get 1 efacten]= [evar] axa ase, & +1) ote | @+tp |, + Dp Usaha. Recurrence Relation for the Probabilities of Binomial Distribution (Fitting of Binomial Distribution). pix+1 (enn) ge Wehave Poy Soa Raat ar (On simplification) _—. el 3) eee perna{ ee 4 pe ... (8-15) which is Fecurrence formula for the probabilities of binomial distribution. This formula provides us a very convenient method of graduating the given data by a binomial distribution. The only probability we need to calculate is p(0) which is given by p (0) = q", where q is estimated from the given data by equating the mean ¥ of the distribution to np, the mean of the binomial distribution. Thus p =X /n. The remaining probabilities, viz., p(1), p(2), ... can now be easily obtained from (8-15) as explained below : n- 0-(24-2)_ xp ©, @)=@+Dki=(F57 4) xP PQ) = [p (x + x41 gq). i n-x p x+1q p(x + hea = ( xample 8-21. The following data due to Weldon shows the results of throwing 12 fair dice 4,096 times; a throw of 4, 5, or 6 being called success. ) xp (2), and so on. va. (8-15a) = ‘ | | Pf | | . | | FUNDAMENTALS OF MATHEMATICAL Stargy,.- Fita binomial distribution and find the expected frequencies. Solution. In the usual notations, we are given : 1 p=Probability of success, ie,, throw of 4,Sor6=3 => 4= 1-p=3- Thus, by the Binomial probability law, the probability of x successes in a throw of 12 dice is given by : P(X=x)=p(x)=( 2) ( Hence in 4,096 throws of 12 dice, the frequencies of x successes are given by : flo) = N. p(x) = 4096. gaz ( Thus the expected frequencies are as tabulated below : COMPUTATION OF EXPECTED FREQUENCIES wx al 12 a ) ‘Success @) Expected Frequency ‘Success @), ZB. 0 1 7 8 Total Eyample 8-22. Seven coins are tossed and number 128 times and the following distribution is obtained : No. of heads Frequencies - 30 n=12,N=4,096 Expected Frequency of heads noted. The experiment is

You might also like