Distribution of the sample mean

75 Views Asked by At

It is my understanding that when you want to find the distribution of the sample mean of some rv, you can do the following.

Let $f(x)$ denote the pdf of $X$:

$$f(x)=\frac{e^{-x (\lambda +\mu )} I_1\left(2 x \sqrt{\lambda \mu }\right)}{\sqrt{\rho } x}$$

Then the mgf of the sample mean $\bar{X}$ is given by $\left[E\left(e^{\frac{t\ X}{n}}\right)]\right]^n$.

Using the above to compute the mgf, I get

$$ M_X(t) =2^{-n} \left(\frac{\lambda +\mu -\sqrt{\left(\lambda +\mu -\frac{t}{n}\right)^2-4 \lambda \mu }-\frac{t}{n}}{\sqrt{\lambda \mu \rho }}\right)^n $$

But when I take the derivative and find the moments (i.e., the mean), it turns out that no matter what the sample size of $n$ is, the sample mean, $\bar{X}$, is exactly the same as the mean of the of the pdf of $X$.

$$E[X] =\frac{2^{-n} \left(\frac{-\sqrt{(\lambda -\mu )^2}+\lambda +\mu }{\sqrt{\lambda \mu \rho }}\right)^n}{\sqrt{(\lambda -\mu )^2}}$$

So I'm confused. I've simulated this rv and I get different results.

So, the question is that is my answer correct? Or did I make an error.

Thanks.

1

There are 1 best solutions below

0
On BEST ANSWER

If $X_1, \ldots, X_n$ are drawn from the same distribution with mean $\mu$, then $$E[\overline{X}] = E\left[\frac{1}{n} \sum_{i=1}^n X_i \right] = \frac{1}{n} \sum_{i=1}^n E[X_i] = \mu,$$ regardless of the value of $n$, and regardless of what the original distribution was.