Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It only takes a minute to sign up.
Sign up to join this community
I have a question about how to derive the distribution of the quotient of two random gamma variables drawn from two different Gamma distributions with the same shape, but different rates. For example, given $$ \theta \sim \frac{1}{Gamma(a, c_1)} \\ \tau \sim \frac{1}{Gamma(b, c_2)} $$ How do I find the distribution of the following? $$ M = \frac{\tau}{\sigma + \tau} $$
I've seen online that it would be $$M \sim \frac{1}{Beta(a+b,c)}$$ if $c_1 = c_2$. But what if $c_1 \neq c_2$?
If $X \sim Gamma(a,1)$ is independent of $Y \sim Gamma(b,1)$ then the ratio $X/Y$ has the Beta prime distribution with parameters $a$ and $b$.
In fact, the result holds if you replace the common value of the rate parameter ($1$ here) by any other value, because the rate parameter has this property: if $X \sim Gamma(a,S)$, then $\lambda \times X \sim Gamma(a, S/\lambda)$ for any $\lambda >0$.
Thus, $X' \sim Gamma(a,c_1)$ is independent of $Y' \sim Gamma(b,c_2)$, then the ratio $X'/Y'$ has the same distribution than $\frac{c_2}{c_1} \times X/Y$ where $X \sim Gamma(a,1)$ is independent of $Y \sim Gamma(b,1)$. Therefore, denoting by $f$ the pdf of the Beta prime distribution then the pdf of $X'/Y'$ is $r \mapsto \frac{c_1}{c_2} f(\frac{c_1}{c_2} r)$. This scaled Beta prime distribution has no devoted name.
