Normalize generalised normal distribution

206 Views Asked by At

Could you please give a hint how to prove that the generalised normal distribution is normalised?

\begin{align} p(x\mid\mu,\alpha,\beta) = \frac{\beta}{2\alpha\Gamma(\frac{1}{\beta})}e^{-(|x-\mu|/\alpha)^\beta} \end{align}

Such that:

\begin{align} \int_{-\infty}^{+\infty} p(x\mid\mu,\alpha,\beta)\,dx = 1 \end{align}

1

There are 1 best solutions below

0
On BEST ANSWER

Without the absolute value, this is often called a Generalized Gamma distribution, which gives a hint to the solution: a Gamma variate $X$ has been shifted to a mean $\mu$ (which will not change the normalization constant), rescaled by $\alpha$ (which divides the normalization constant by $\alpha$, as you can see), and then takes the $\beta \gt 0$ power. The absolute value introduces a mirror image of the distribution on the negative reals, whence we need to divide by $2$.

That shows we only need to check that the integral

$$I(\beta)=\int_0^\infty \exp(-u^\beta) du$$

equals $\frac{1}{\beta} \Gamma\left(\frac{1}{\beta}\right)$ and strongly suggests the substitution $v=u^\beta$, whence $$du = d(v^{1/\beta}) = \frac{1}{\beta}v^{1/\beta-1}dv$$ and

$$I(\beta) = \int_0^\infty \frac{1}{\beta}v^{1/\beta-1} e^{-v}dv = \frac{1}{\beta} \Gamma\left(\frac{1}{\beta}\right),$$

QED.