Generalized Gaussian distribution

From Wikipedia, the free encyclopedia

[edit] Generalized Gaussian Distribution (GGD)

A random variable X has generalized Gaussian distribution if its probability density function (pdf) is given by

f(x;m,\sigma,\alpha)=a\ \exp(-|(x-m)/b|^\alpha) ,x \in \R,
where m is the mean of the distribution, σ is the standard deviation, α is the shape parameter and σ,α > 0. a and b are computed according to :

a = \frac{1}{2\Gamma(1+1/\alpha)b }, \quad
b = \sigma\,\sqrt{\frac{\Gamma(1/\alpha)}{\Gamma(3/\alpha)}}.

b is a scaling factor which allows the variance to be σ2. \Gamma(\cdot) stands for the Gamma function.

When α = 1 , f(x;m,σ,α) corresponds to a Laplacian or double exponential distribution, α = 2 corresponds to a Gaussian distribution, whereas in the limiting cases where α approaches +\infty the pdf ( f(x;m,σ,α) ) converges to a uniform distribution in (m-\sqrt{3}\sigma, m+\sqrt{3}\sigma). When \alpha \rightarrow 0^+ a degenerate distribution in x = μ is obtained.

As GGD is symmetric about its mean (m), odd-order central moments are zero.[1]

[edit] References

  1. ^ J. Armando Domínguez-Molina, Graciela González-Farías,Ramón M. Rodríguez-Dagnino,"A practical procedure to estimate the shape parameter in the generalized Gaussian distribution".

2.Varanasi, M.K., Aazhang, B. (1989). Parametric generalized Gaussian density estimation,J. Acoust. Soc. Am. 86 (4), October 1989, pp. 1404.