Transformed gamma distribution

122 Views Asked by At

f(x) is the pdf of two parameter gamma distribution. Gamma(alpha, beta).

$$f(x)=\frac{x^{\beta-1}*exp(-x/\alpha)}{(\alpha^\beta) *\Gamma(\beta)}$$

x>0

I need to transform x as

y=(x-mu)/sigma, x=mu+y*sigma, dx/dy=sigma

f(y)=[sigma*((mu+y*sigma)^(beta-1))exp{-(mu+ysigma)/alpha}]/{(alpha^beta) *gamma function(beta)}

1

There are 1 best solutions below

1
On

Your density looks the density of a $\text{Gamma}(\beta;\alpha)$

If you want a density of a $\text{Gamma}(\alpha;\beta)$ this is the following

$$f(x)=\frac{x^{\alpha-1}\cdot \text{exp}(-x/\beta)}{\beta^\alpha\cdot \Gamma(\alpha)}$$

with

$$E(X)=\alpha\cdot\beta$$

$$V(X)=\alpha\beta^2$$

Thus using your transformation function

$$Y=\frac{X-\alpha\beta}{\beta\sqrt{\alpha}}$$

where

$$x=\beta\sqrt{\alpha}y+\alpha\beta$$

$$x'=\beta\sqrt{\alpha}$$

and thus

$$f_Y(y)=\frac{\beta\sqrt{\alpha}}{\Gamma(\alpha)\beta^\alpha}(\beta\sqrt{\alpha}y+\alpha\beta)^{\alpha-1}\text{exp}\left\{-\frac{\beta\sqrt{\alpha}y+\alpha\beta}{\beta}\right\}$$

simplifying...

$$f_Y(y)=\frac{(\sqrt{\alpha})^\alpha}{\Gamma(\alpha)}(y+\sqrt{\alpha})^{\alpha-1}\text{exp}\left\{-\sqrt{\alpha}(y+\sqrt{\alpha})\right\}\cdot\mathbb{1}_{[-\sqrt{\alpha};+\infty)}(y)$$