Scribd
Upload
15 views57 pages

MGF CHF

Moment generating function notes

Uploaded by

lakshaaysharma003
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
15 views57 pages

MGF CHF

Moment generating function notes

Uploaded by

lakshaaysharma003
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
The moment generating function (m.g.f.) of a random variable X is denoted by Mx (t) and is defined as Mx(t)=E(e™), tER wherever this expectation exists. For discrete random variable For continuous random variable, My(t) = Yerre@ Mx(t) = * tx f(x) dx Where p(x) is the probability Where f (x) is the probability mass function (p.m.f.) density function (p.d.f.) — — ay As its name implies, the m.g.f. can be used to compute a distributions moments: othe némoment about O is then‘! derivative of the moment-generating function, evaluated at 0. s However 1 all random variables have moment-generating functions. lee Example: Let X have the probability mass function 61 f@ = fesse’ x= 1, 0; otherwise Find the moment generating function of X. Solution: By definition, we have Mx(t) = E(e™) tx ox Here, the series > is divergent x1 (not convergent) by D’Alembert's Ratio Test for every. t > 0. Hence, the m.g.f. of X DOES NOT EXIST. — Example: Find the m.g.f of the Poisson distribution. Does My (t) exist for all values of t? Solution: The p.m.f. of the Poisson distribution with parameter 2 is -Agx p(x) = nO Thus, m.g.f. of it is eae S tx 2 OA Mx(t) = I eta)y* x! 5x = 0,1,2,... Thus, My (t) = e7Aa-e') which exist for all values e 2 yn Because, e* =) — and n=o n! is convergent for alLx. ee For any positive integer r, we denote H, = E(X") 1" Method 2" method: r da H, = Coefficient of Sin the series expansion Br = a (Mx) att=0 ” r of Mx(t) For more details, Mean = E(X) My(t) = E(e'*) 2 =E ome Ede ore =eli+ix tex + Variance = E(X?) — [E(X)]? t? =1+0t8(X) + FO) Ft ccs 2 Hltintting ‘ie Example: Find the m.g.f of the Poisson distribution and hence fina its mean and Variance) Does My (t) exist for all values of t? Ww Find Mean and Variance: Mx(t) = e7Ala-e') a ‘ =, ee BQ) = 5 [eR] MEH 0, y2yaata = eAlti-eVQet) Therefore, = Mean = E(X) E(X2) = Gale te) =lA” = fet“ 208] Variance = E(X?) — [E(X)]* =A+2-2 = eAlt-e') (get) + e~AG-e) get)? =A N oo Example: Find the m.g.f of the < distribution and hence find its mean and Vari ON Example: Find the m.g.f. for the geometric distribution and find its Solution: The p.m.f. of the geometric distribution is p@) =p 5 x= 12)... Thus, m.g.f. is given as My(t) = E(e"*) Yeap mean and variance. (8 Characteristi Example: A perfect coin is tossed twice. Find the m.a.t. of the umber of heads. Hence, find the mean andvarionce, Solution: The p.m.f, of X, (the number of heads) is The m.g.f. ic x 0 1 2 Mx(t) = E(e) P(x) % % ‘% 2 Gee Gee vBinesiol _1+2e' +e n mci (pe +4) i 4 2 = (bee uy 7 — ¢ SS atest ys a sane EY ee a E(X) = qi z EO?) = my) Lge + 2e74) Mean and variance: Att = 0, we have BQ)=1; Y gx?) =3/2 iene Example: Find the m.g.f. for the distribution where a fixe = 1 f@) oe 0 ; otherwise Solution: Since X Is a discrete random variable with points | and 2 only. EW). ewe £ The m.g.f. is My(t) = De" f(x) -<@-e@ 2et + et 3 3, Example: The m.g.t. of a random variable X Is glven by My(t) = e3(¢*-1), Find P(X = 1). Solution: Since, the given m.g.f. can be rewritten as Mg(t) = e-30-#*) Thus, by comparing It with m.g.f. of Polsson distribution, My(t) = e-AG-et) SNS Therefore, by Poisson distribution weget, 229 hall Example: Find the m.g.f. for the exponential density function and hence calculate the me and variance. Solution: The p.d.f. for the exponential distribution Is _fae*; x20 f @=ff ; otherwise By definition, we have e-A-ox\” =2 My(t) = E(e™) (S - 5), | Je" dx =a(o+ 1 ) Provided 2—t > 0 fo A-t be, t > > Be etb_eta (b-a)t Mx(t) = Thus, EQ) = = MeO $term . containing "t' EX?) = + term. containing "t" Smo Example: (Laplace distribution) If X has the probability density function F(x) = $e“, —co < x < 00, Find the mean and variance of X. Solution: The p.d.f of the Laplace distribution is *; x>0 Z 5e f@= 47 zen x<0 et = ={ eter ax | e™ =e dx 00 2 0 2 1/° 1” =3] e(ttx dx+3f ef DF dx co 0 tfettDe\® 4 /el(t-D\? =3(saz)_ *2(T7), 1 2 Providedt +1 >0 and t-1<0 ie,¢>—landt<1 ie, -1 ef é - a ap" sr Yee (« = 1) er xer © = (pet) > ¢ 7 i @er er = (pety” > € +r- 1) (aety* r-1 kao roth =pre' Properties of the ‘Sum of a negative binomial series G-or=y hen) n-1 = Wet 1 gety Provided qe] <1 => t Ele 02) Ef -ab : -e é if (¢ ) = Bh 2 ele ) - 2°, mH) : "lk G sy] Example: If the m.g.f. of X is given by My (t) = za + e*)?, then find the mg.f. of Y = Solution: Since ma.f. of X is 1 My) = 7 + et)? Thus, by using the property, Therefore, at Mx-a(t) = e7 2M, s=4(0) = e77 My (5) ¢ Mx-o(t) = 7 My (). : we get Example: If the mast. of X is given by. My(t) = >-(e% — 1)? , then find the mgt. of ¥ = = Solution: Since m.g.f. of X is My(t) = ze -1)? Thus, by using the property, aa : Mszo(t) = e7* My(;), we get Example: If the m.g.f. of X is given by My (t) = = , [t] <1, then find the mg.f. of Y = Solution: Since mast. of X is My(t) == Therefore, Thus, by using the property, ot t Mx-a(t) = e7% My (=), 2 ) x (i) we get ] Uniqueness theorem of M.G.F. The moment generating function of a distribution, if it exists, snipialy erect: the distribution. This means that corresponding to a given probability distribution, there is only one m.g.f. (if it exists) and corresponding to a given m.g.f., there is only one probability distribution. Hence, My(t) = My(t) => X &Y are identically distributed. 7 r Lv + ae ¥ ot i wy laefec Le fol Characteristic function 2 ip fae = 4 The characteristic function of a random variable X is denoted by y(t) and is defined as P24 x(t) =E(e"), tER lox) < J For discrete random variable $x) = Yep) For continuous random variable, éx(0 { if “ety ay} fl 4 Where (1x) is the probability density function (p.d.f.) * Where p(x) is the probability mass function (p.m.f.) For any positive integer 1, we denote Hy = E(X") 1" Method 1a" B= apr (bx) aee=0 y(t) = ee) 242 2" method: =e [rent as . . Pe Hy = Coefficient of the series =1 + EQ) +P EO) ++ expansion of y(t) : =1 +g em Ge wy eel: inte caer Tne ges bution and hence find its mean and Variance, Solution: Meio wy Example: Find the characteristic function of the Binomial distribution and hence find its mean and Variance. Solution: m= aes w Example: Find the characteristic function for the geometric distribution and find Its mean and variance, Solution: The punt. of the geometric distribution ts PQ) =a" tps x= 12, ee ‘Thus, characterictic function Is given as x(t) = E(e**) w rane: hpertet xin i tossed ie Find he characteris fntion othe BBE Wa) Hence, find the mean and variance. a Ibs . Con —> Hite . e) Ly wtih Dist Bimorniat_ Qn ws s vu 3 Wwe. Ce2)s) Vase i b SS ww Example: A perfect coin is tossed twice. Find the characteristic function of the lumber OF heads. Hence, find the mean and variance. Solution: The p.m.t. of X, (the number of heads) is = 0 = pa) %& Ta ) The characteristic function is ede | = (et + 202)] $x = E(e*) 2 = Lel*p(x) 1 1 =09(;) +e" ()+e*"( \ 4, 2, Mean = we it 4 o2te sitet’ Variance = E(X2) — [EOD 1 2 wy Example: If characteristic function of a random variable X is given as a 4 leit 41,31 Ox =Z4 Fe te Find P(X < 1). ‘Solution: By definition of characteristic function ox) = E(e*) 1.1 1 Titles ¥ ot aptzel tie x p(x) 7d S Gt get + ze = pl) + eltp(t) + ep) + On comparing, we get PO=2 : p=F + pey=o: P@=% + pa) =P6)= w Example: Find the characteristic function 2 si ox=l 3 for the distribution f@=s7 1 3 x=2 w 0 ; otherwise Solution: Since X is a discrete random variable with points | and Z only. The characteristic function is Ox) = Le f(x) “QoQ 2elt + eit 3 x Property I: The characteristic function y(t) Is bounded 9) Ihe, 1x(O) S 1 for all ¢ Proof: Let X is a continuous random variable. By definition of characteristic function, we have losf0l = |E(e™)| J e!™ F(x) dx| *» < [roa s fiesaiax ” <1 because f(x) Is a pdf. Hence, < J elf (x) dx lox(t)| <1 forall t IA Property I: The characteristic function q(t) Is bounded by I, Le., Iby(t)| < 1 for all ¢ Proof: Similarly, for discrete random variable, we get lox() = [E(e"*)] = Ir efx) = le reo =¥reo <1 because f(x) Isa [Link]. Remarks: The characteristic function dy (t) exists for all value of t. as Property 2: For a characteristic function cby (t), we have by (0) = 1 Proof: By definition of by (t), we have x (0) = E(e*) = E(1) =ly aa Property 3: y(t) and by(—t) are conjugate functions, Le., y(t) = ox(-t) Proof: By definition of y(t), we have Ox = E(e"™*) x(t) = E(e'™*) Property 4: If a random variable X has a symmetrical distribution about the origin, Le., f(-x) = f(x) then y(t) Is a real-valued and an even function of . Proof: We prove this result only for the continuous case. (The discrete case can be solved similarly) ox) = pI | el* F(x) dx . e i ef(-y) dy For real-valued function, we have $x = dx(-O = xO =6 J e-"F(y) dy which shows that y(t) is real-valued. = ox(-t) ) Hence, Py is an even function of t. Property 5: If X, and X2 are independent random variable then $x,+x, 0 = x, 06x, Proof: By definition of characteristic function, we have Pxx0(0 = E(ett%+42)) = E(citXstit%2 ) 7 E(eit*eitXy Because, if X & Y are independent 7 E(e!*) Ee) random variables, then E(XY) = E(X)E(Y), while converse is not true = bx, Obx,0 i Remarks: a In general, if X;,i=1,2,...,n be an independent random variables, then Sayexpename(O =| [bx i=1 ie., characteristic function of the sum of the random variables is equal to the product of the characteristic function of the variables. Effect of Change of origin and scale Property 6: If we transform the variable X to a new variable Y by changing both the origin and the scale 7 _iat t inX,be,¥==then y(t) =e F by () Proof: By definition of characteristic function, we have y(t) = E(e”) it(x-a) ) itx =E(en “é x =e"NE (") =e dy (6) Result: dx is uniformly continuous in R. q Cicer) ecu Property 7: If X is a random variable with characteristic function dy (t) and if. uw, = E(X"), : then = = My = apr hx at t= 0 Proof: We prove this result only for the continuous case. (The discrete case can be solved similarly) 1 i (in) O=— [CFe ax é 2a Je a Ui Oy Differentiating under the integral sign 7 times with respect to t, we get seo) = J (ixyel™f(x)dx Uniqueness theorem of p(t). The characteristic function uniquely determines the distribution function. 7a In other words, if X & Y are any random variables, then ox® = by) & Fy = Fy Inversion Theorem Let X be a continuous random variable with the distribution function Fy, p.d.f., f (X) and characteristic function py (t), then p.d.f. f(x) can be expressed in terms of y(t) by wef k fray eso few Gyltat dx an _ =o

You might also like