Question

 Let Xn Pois (n) for all positive integers n . Use MGFs to show that the distribution of  the standardized version of Xn converges to a Normal distribution as n, without  invoking the CLT. \begin{array} { l } { \text { Let } X _ { n } \sim \text { Pois } ( n ) \text { for all positive integers } n \text { . Use MGFs to show that the distribution of } } \\ { \text { the standardized version of } X _ { n } \text { converges to a Normal distribution as } n \rightarrow \infty , \text { without } } \\ { \text { invoking the CLT. } } \end{array}

Solution

Verified
Step 1
1 of 2

We are given that XnPois(n)X_n \sim \text{Pois}(n). Since E(Xn)=nE(X_n) = n and Var(Xn)=n\operatorname{Var}(X_n) = n, we define standardized variables

Yn:=Xnnn\begin{align*} Y_n := \frac{X_n - n}{\sqrt{n}} \end{align*}

Let's find the MGF of YnY_n. We have that

MYn(t)=E[etYn]=E[etXnnn]=etnMXn(tn)\begin{align*} M_{Y_n}(t) = E[e^{tY_n}] = E[e^{t\frac{X_n - n}{\sqrt{n}}}] = e^{-t\sqrt{n}}M_{X_n}(\frac{t}{\sqrt{n}}) \end{align*}

Since we know MGF of Poisson distribution, we have that

etnMXn(tn)=etnen(et/n1)\begin{align*} e^{-t\sqrt{n}}M_{X_n}(\frac{t}{\sqrt{n}}) = e^{-t\sqrt{n}} e^{n(e^{t/\sqrt{n}}-1)} \end{align*}

We are interested in what happens with MYn(t)M_{Y_n}(t) as nn \rightarrow \infty. Let's consider the expression that is in exponential function. We have that

limn(n(et/n1)tn)=limx(x(et/x1)tx)=[use x=1y2]=limy0(1y2(ety1)t1y)=limy0ety1tyy2=[L’Hopital]=t2limy0ety1y=t22\begin{align*} \lim_{n \rightarrow \infty} \left(n(e^{t/\sqrt{n}}-1) -t\sqrt{n} \right) &= \lim_{x \rightarrow \infty} \left(x(e^{t/\sqrt{x}}-1) -t\sqrt{x} \right) = [\text{use } x = \frac{1}{y^2}] \\ &= \lim_{y \rightarrow 0} \left(\frac{1}{y^2}(e^{ty}-1) -t\frac{1}{y} \right) \\ &= \lim_{y \rightarrow 0} \frac{e^{ty}-1-ty}{y^2} = [\text{L'Hopital}] \\ &= \frac{t}{2} \lim_{y \rightarrow 0} \frac{e^{ty}-1}{y} = - \frac{t^2}{2} \end{align*}

Hence, we have that

MYn(t)=etnen(et/n1)et22\begin{align*} M_{Y_n}(t) = e^{-t\sqrt{n}} e^{n(e^{t/\sqrt{n}}-1)} \rightarrow e^{- \frac{t^2}{2}} \end{align*}

as nn \rightarrow \infty. But the limit MGF is exactly the MGF of Standard Normal Distribution. Hence

YnN(0,1)\begin{align*} Y_n \rightarrow \mathcal{N}(0,1) \end{align*}

Create a free account to view solutions

Create a free account to view solutions

Recommended textbook solutions

Probability and Statistics for Engineers and Scientists 9th Edition by Keying E. Ye, Raymond H. Myers, Ronald E. Walpole, Sharon L. Myers

Probability and Statistics for Engineers and Scientists

9th EditionISBN: 9780321629111 (1 more)Keying E. Ye, Raymond H. Myers, Ronald E. Walpole, Sharon L. Myers
1,204 solutions
Introduction to Probability 2nd Edition by Jessica Hwang, Joseph K. Blitzstein

Introduction to Probability

2nd EditionISBN: 9781138369917Jessica Hwang, Joseph K. Blitzstein
621 solutions
Probability and Statistics for Engineering and the Sciences 9th Edition by Jay L. Devore

Probability and Statistics for Engineering and the Sciences

9th EditionISBN: 9781305251809Jay L. Devore
1,589 solutions
Statistics and Probability with Applications 3rd Edition by Daren S. Starnes, Josh Tabor

Statistics and Probability with Applications

3rd EditionISBN: 9781464122163Daren S. Starnes, Josh Tabor
2,555 solutions

More related questions

1/4

1/7