6
$\begingroup$

$Y$ is a random variable with $$M(t) = \frac{1}{(2-\exp(t))^s}.$$

Does $$\frac{Y-E(Y)}{\sqrt{\operatorname{Var}(Y)}}$$ converge in distribution as $s$ tends to infinity?

I let $Z = \frac{Y-E(Y)}{\sqrt{\operatorname{Var}(Y)}}$. Differentiating the MGF of $Y$ and let $t = 0$ we have $$E(Y) = s,$$$$E(Y^2) = s^2 +2s,$$$$\operatorname{Var}(Y) = 2s.$$ Thus Mgf of Z: \begin{align*}E(\exp(tz)) &= E\left(\exp\left(\frac{t(y-s)}{\sqrt{2s}}\right)\right)\\ &=E\left(\exp\left(\frac{ty}{\sqrt{2s}}\right)\exp\left(\frac{-ts}{\sqrt{2s}}\right)\right)\\ &=E\left(\exp\left(\frac{ty}{\sqrt{2s}}\right)\exp\left(\frac{-ts}{\sqrt{2s}}\right)\right)\\ &=\exp\left(\frac{-ts}{\sqrt{2s}}\right)\frac{1}{\left(2-\exp\left(\frac{t}{\sqrt{2s}}\right)\right)^s}. \end{align*} But I'm not sure how to proceed - seems to me it tends to $0$?

$\endgroup$
1
$\begingroup$

As you have determined: $$ M_Z(t) = \mathbb{E}(\exp(t Z ) ) = \frac{\exp\left(-t \frac{s}{\sqrt{2s}}\right)}{\left( 2 - \exp\left(\frac{t}{\sqrt{2s}} \right) \right)^s} = \left( \frac{\exp\left(-t \frac{1}{\sqrt{2s}}\right)}{ 2 - \exp\left(\frac{t}{\sqrt{2s}} \right) } \right)^s $$ In order to determine large $s$ behavior, use Taylor series: $$ \frac{\exp\left(-t \frac{1}{\sqrt{2s}}\right)}{ 2 - \exp\left(\frac{t}{\sqrt{2s}} \right) } = \frac{1- \frac{t}{\sqrt{2s}} + \frac{t^2}{4 s} + o\left(\frac{t^2}{s}\right)}{ 1- \frac{t}{\sqrt{2s}} - \frac{t^2}{4 s} + o\left(\frac{t^2}{s}\right) } = 1 + \frac{t^2}{2s} + o\left(\frac{t^2}{s}\right) $$ Thus $$ M_Z(t) = \left( 1 + \frac{t^2}{2s} + o\left(\frac{t^2}{s}\right) \right)^s $$ The large $s$ limit is now easy to find: $$ \lim_{s\to \infty} M_Z(t) = \exp\left( \frac{t^2}{2} \right) $$ This is the mgf of the standard normal variable, as expected by CLT.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.