If I have a variable $X$ that has a gamma distribution with parameters $s$ and $\lambda$, what is its momment generating function.

I know that it is $\int_0^\infty e^{tx}\frac{1}{\Gamma(s)}\lambda^sx^{s-1} e^{-x\lambda}dx$ and the final answer should be $(\frac{\lambda}{\lambda-t})^s$, but how can i compute this?

P.S. I know that there are other questions on this site about the MGF of the gamma distibution, but none of those use this specific definition for the density function of a gamma distribution. And I would like to see it with this one. Thanks a lot!

  • $\begingroup$ Look at Wikipedia. It has such info for most of the common probability distribution families. Usually, in a box on RHS very near the top. $\endgroup$ – BruceET Apr 21 '15 at 20:28

Your question ultimately becomes to show the following: $$\int_0^\infty e^{tx}\frac{1}{\Gamma(s)}\lambda^sx^{s-1} e^{-x\lambda}\ dx = \left(\frac{\lambda}{\lambda - t}\right)^s$$ $$\int_0^\infty e^{tx}\frac{1}{\Gamma(s)}\lambda^sx^{s-1} e^{-x\lambda}\ dx = \frac{\lambda^s}{\Gamma(s)}\int_0^\infty x^{s-1}e^{-(\lambda-t)x} \ dx$$ From repeated applications of integration by parts, it is known that $$\int_0^\infty x^ae^{-bx} \ dx = \frac{\Gamma(a+1)}{b^{a+1}}$$ Making this substitution we get: $$\frac{\lambda^s}{\Gamma(s)}\int_0^\infty x^{s-1}e^{-(\lambda-t)x} \ dx = \frac{\lambda^s}{\Gamma(s)}\frac{\Gamma(s)}{(\lambda - t)^s} = \left(\frac{\lambda}{\lambda - t}\right)^s$$


I'd like to give another answer. Instead of the "repeated integration by parts" in the other answer, we can do the following:

We know the definition of the gamma function to be as follows:

$$\Gamma(s) = \int_{0}^\infty x^{s-1}e^{-x}dx$$

Now $\int_0^\infty e^{tx}\frac{1}{\Gamma(s)}\lambda^sx^{s-1} e^{-x\lambda}dx$ = $\frac{\lambda^s}{\Gamma(s)}\int_0^\infty e^{(t-\lambda)x}x^{s-1}dx$. We then integrate by substitution, using $u = (\lambda - t)x$, so also $x=\frac{u}{\lambda-t}$. This gives us $\frac{du}{dx}=\lambda - t$, i.e. $dx = \frac{du}{\lambda-t}$. Now let's put this into the integral, so then we get:

$$\frac{\lambda^s}{\Gamma(s)}\int_0^\infty e^{-u}{\left(\frac{u}{\lambda -t}\right)}^{s-1}\frac{du}{\lambda-t} = \left(\frac{\lambda}{\lambda-t}\right)^s\frac{1}{\Gamma(s)}\int_0^\infty u^{s-1}e^{-u}du$$

Here on the right-hand side we recognize the integral as the gamma function, so we get $\left(\frac{\lambda}{\lambda-t}\right)^s\frac{\Gamma(s)}{\Gamma(s)}$. Those gamma's cancel so then we get what we want:



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.