MGF from a conditional distribution

993 Views Asked by At

I'm having a difficulty with this problem, it is not for submission.

Let Y~U(0,1) be a continuous random variable, and X is another random variable such that given Y=p: $X|_{Y=p}$~B(n,p).
I need to find the moment generating function of X.
I honestly don't have much here, but I will try to add some context. It's weird to me that they ask about Y=p while being a continuous random variable, but my guess is that this will be relevant later.
I have the MGF formula of the Unified distribution: $$M_x(t) = \frac{e^{tb}-e^{ta}}{t\left(b-a\right)} $$ And I thought about using that, but I still don't really see the way. Another direction I thought of going through but couldn't advance with it is going to the definition: $$M_x(t) = \sum _x\:e^{tx}P\left(X_{|Y=p}=x\right) $$ but I couldn't really solve it.
Thanks in advance for any help provided.

2

There are 2 best solutions below

2
On BEST ANSWER

We can first write conditional MGF of bernoulli random r.v.$(I\sim BN(p))$, $MGF_{I}=pe^{t}+1-p$ then given that binomial r.v. is $B(n,p)=\sum_{k=1}^{n}I_{k}$ the MGF of binomial r.v. $B(n,p)$ is, \begin{equation} M(p,n)=E(e^{t\sum_{j=1}^{n}I_{j}})=\Pi_{k=1}^{n}E\left[e^{tI_{k}}\right]=\left(pe^{t}+1-p\right)^{n}=\left(p(e^{t}-1)+1\right)^{n} \end{equation}

Given that $p=Y$, \begin{equation} M(X,t,n,p=y)=\left(y(e^{t}-1)+1\right)^{n} \end{equation}

Now given that $p=Y\sim U(0,1) $ we can compute the MGF as follows, \begin{equation} M(X,n,p=Y\sim U(0,1))=\int_{0}^{1}\left(y(e^{t}-1)+1\right)^{n}dy \end{equation}

Using standard integration arguments for above, we will have the final MGF, \begin{equation} M(X,t,n)=\frac{e^{t(n+1)}-1}{(n+1)(e^{t}-1)}=\frac{\sum_{k=1}^{n}e^{t(n-k)}}{n+1} \end{equation} A quick check on first moment will be which is, \begin{equation} E\left(X(n,y)\right)=nE(y)=\frac{n}{2} \end{equation} \begin{equation} \frac{\partial M(X,t,n)}{\partial t}|_{t=0}=\frac{1}{n+1}\sum_{k=1}^{n}(n+1-k)=\frac{n(n+1)}{2(n+1)}=\frac{n}{2} \end{equation} which confirms the MGF.

Hope this is helpful for your question.

0
On

Starting with the definition is the right step. Insert in the sum the expression for $B(x,n,p)$ ,considering for the moment $p$ to be fixed: you'll get a $M_x(t,p,n)$, that is the mgf subject to being $Y=p$. Thereafter, multiply that for $Pr(Y=[p, p+dp))$ and integrate over the support of $Y$