finding inverse gamma distribution in bayesian analysis

hello, I have been working with hyperspectral images.Actually i am trying to make use of hierarchical bayesian model to perform unmixing of the image.In the process i need to find the noise variance.this variance is the inverse gamma distribution of 2 parameters. one of them is a constant having a value of about 77 and the other parameter has a quite high value,in the range of 10^14.i have been trying to use the gaminv function of matlab, but it has not been of any use.I am getting NaN as my answer,dont knw why..?Plz help me out....

thanks, Ekta

Answer 1

[Roger Stafford]

It looks as though the definition of gamma distribution parameters you are using does not coincide with that used by the matlab 'gaminv'. I suggest you carefully read the definition given at:

 http://www.mathworks.com/help/stats/gaminv.html

and adjust your own parameters accordingly if you wish to use 'gaminv'. You can read about three different kinds of parameter definitions in common use in the Wikipedia website

 http://en.wikipedia.org/wiki/Gamma_distribution.

The parameters used by matlab are those referred to in this latter article as "shape k" and "scale theta". In Bayesian statistics you are more likely to be using the reciprocal of the scale parameter and be calling it the rate parameter.

[Ekta Shah]

Thank you for the suggestion.I shall try it out and get you know...Thank You....