0
$\begingroup$

Is the following derivation for scaling a Generalized Gamma distribution correct?

Given $X\sim GG(x;a,d,p)$, with $x\ge 0$, $a,d,p > 0$ and pdf

$$f_x(x;a,d,p) = \frac{px^{d-1}exp\Big(-(x/a)^p\Big)}{a^d\Gamma(d/p)}$$

Let $Y=g(X)=kX$ and, consequently, $X=g^{-1}(Y)=Y/k$, and the Jacobian $dX/dY=1/k$. Then, the transformation formula holds:

$$f_y(y)=f_x(g^{-1}(y))\left|\frac{dx}{dy}\right|$$

Hence we have

$$f_y(y;ka,d,p) = \frac{py^{d-1}exp\Big(-(\frac{x}{ka})^p\Big)}{(ka)^d\Gamma(d/p)}$$

$\endgroup$
0
$\begingroup$

The general approach is correct but you have mistakenly written $\exp(-(\frac{x}{ka})^p)$ rather than $\exp(-(\frac{y}{ka})^p)$. The conclusion is that if $X$ is generalized with parameters $a$, $d$, $p$ (which are scale, shape, and "power," respectively), then $Y = kX$ is generalized with scale $ka$, with the same shape and power parameters. This makes sense since this is consistent with the notion of a scale parameter.

$\endgroup$

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.