[SciPy-user] multivariate distributions

Christopher Fonnesbeck fonnesbeck at mac.com
Wed Jun 2 08:01:48 CDT 2004


On Jun 2, 2004, at 1:57 AM, Scott Bray wrote:

> G'day Robert,
>
> i was kinda hoping you'd reply - you helped me out a little while ago. 
> well my project's sinced progressed and i've hit another problem. hope 
> you dont mind if i pick your brain again...
>
> i've done some wide reading and have found lots of info on multivarite 
> normal dists and some equations for the joint density of a 
> multivariate normal, but it doesn't seem useable to me. i have 5 lots 
> of N observations but the multinormal needs means and covariances 
> which i have been calculating FROM the functions (i did this using the 
> cauchy). i have heard about multivariate cauchy's but haven't been 
> able to find any concrete info.
>
> so my question is - how do i 'derive the formula on paper first'?


I'm not sure how a cauchy is going to help you get at a multivariate 
normal distribution. Are your data overdispersed, or something? Most 
statistics texts will give you the pdf of a multivariate normal; you 
just have to code that into python. I have a multivariate normal and 
wishart (i.e. multivariate gamma) programmed in FORTRAN and compiled as 
a python module using f2py, if you want it.

The rv module in scipy.stats has a multivariate_normal function, but 
this simply generates random variables.

C.

--
Christopher J. Fonnesbeck ( c h r i s @ f o n n e s b e c k . o r g )
Georgia Cooperative Fish & Wildlife Research Unit, University of Georgia

"The primary objective of copyright is not to reward  the labor of 
authors, but [t]o promote the Progress of Science and useful Arts. To 
this end, copyright assures authors the right  to their original 
expression, but encourages others to build  freely upon the ideas and 
information conveyed by a work. This  result is neither unfair nor 
unfortunate. It is the means by which  copyright advances the progress 
of science and art."
-- US Supreme Court Justice Sandra Day  O'Connor



More information about the SciPy-user mailing list