I'm looking for a way to compute the CDF (or PDF) of the following random variable:
Let $B\sim Beta(\alpha, \alpha)$ for some $\alpha \geq 1$. Then the random variable in question is $B^2-B$.
Trying to compute the CDF directly from the PDF of the beta distribution results in some nasty integration which I'm not sure how to complete. Ideas would be appreciated!
Let's set
$X\sim Beta(a;a)$ and $Y=X^2-X$
Y is defined on the support $y \in(-0.25;0]$
The transformation function is not monotonic but we can use the Fundamental transformation theorem getting Y density inverting the function as a piecewise one getting
$$f_Y(y)=\frac{\Gamma(2\alpha)}{\Gamma(\alpha)\Gamma(\alpha)}\Bigg[\Bigg(\frac{1}{2}+\sqrt{y+\frac{1}{4}}\Bigg)^{\alpha-1}\Bigg(\frac{1}{2}-\sqrt{y+\frac{1}{4}}\Bigg)^{\alpha-1}+\Bigg(\frac{1}{2}-\sqrt{y+\frac{1}{4}}\Bigg)^{\alpha-1}\Bigg(\frac{1}{2}+\sqrt{y+\frac{1}{4}}\Bigg)^{\alpha-1}\Bigg]\frac{1}{2\sqrt{y+\frac{1}{4}}}$$
$$ \bbox[5px,border:2px solid black] { f_Y(y)=\frac{\Gamma(2\alpha)}{\Gamma(\alpha)\Gamma(\alpha)}\frac{\Bigg(\frac{1}{2}+\sqrt{y+\frac{1}{4}}\Bigg)^{\alpha-1}\Bigg(\frac{1}{2}-\sqrt{y+\frac{1}{4}}\Bigg)^{\alpha-1}}{\sqrt{y+\frac{1}{4}}}\cdot\mathbb{1}_{(-0.25;0]}(y) \ } $$
Examples:
$$f_Y(y)=\frac{1}{\sqrt{y+\frac{1}{4}}}\cdot\mathbb{1}_{(-0.25;0]}(y)$$
$$f_Y(y)=\frac{-6y}{\sqrt{y+\frac{1}{4}}}\cdot\mathbb{1}_{(-0.25;0]}(y)$$
You can easy check that the above expressions are nice densities