Moment generation function v.s. density function.

382 Views Asked by At

I have a curiosity about the relationship between m.g.f. and p.d.f..

Actually, I knew like the following: $$M_X(t)=E\left[e^{tX}\right]$$ where $M_X(t)$ is the moment generating function with respect to random variable $X$.

Also, I learned that m.g.f. is unique. So if m.g.f. of $X$ and $Y$ are same, then $X$ and $Y$ have the same distribution.

However, my professor mentioned slightly that knowing all moments is equivalent to knowing density function.

Under any circumstances, is it possible to make a random variable's density function if I know its moment generating function?

I agree that if I have moment generating function which is well-known like Bernoulli, Poisson, ..., I can guess its density function.

However, I think it is not possible to make density function, even if I have a value of $E[X]$, $E[X^2]$, $E[X^3]$, and so on...

Is my thought wrong?

1

There are 1 best solutions below

0
On BEST ANSWER

The m.g.f. $M(t) = E[e^{tX}]$, if it exists for $t$ in some real interval $-\epsilon < t < \epsilon$, also exists and is analytic for $-\epsilon < \text{Re}(t) < \epsilon$ in the complex plane. In particular, the analytic continuation of the m.g.f. to imaginary $t$ is the characteristic function $C(s) = E[e^{isX}]$. We can then obtain the density, if indeed it exists, as the Fourier transform of the characteristic function.