[SciPy-user] Bug in scipy.stats.rv_discrete is on line 3218

[Michael Nandris]

Stefan, 

I have traced the bug to line 3218 of distributions.py 

The problem is with the operation of argmax in the _drv_ppf handler.

It can be solved by pulling the correct state out of self.Finv (using a probability, as generated by mtrand.random_sample), in an order -preserving manner.

As soon as tracs starts working i'll post some more demo testcode

cheers

Michael



Stefan van der Walt <stefan at sun.ac.za> wrote: Hi Michael

Thanks for the report.  This should be fixed in SVN revision 3246.

Cheers
Stéfan

On Tue, Aug 14, 2007 at 01:08:45PM +0100, Michael Nandris wrote:
> Hi,
> 
> I think there may be an issue with the way rv_discrete orders its output when
> it encounters zeros in the input, causing non-zero probabilities to creep up
> towards the end of the output. If you are attempting to track the accumulation
> of states in an n-state Markov chain, this is a problem!
> 
> I have had a look at the API but can't figure it. Any help at fixing this would
> be much appreciated.
> 
> regards
> 
> M.Nandris
_______________________________________________
SciPy-user mailing list
SciPy-user at scipy.org
http://projects.scipy.org/mailman/listinfo/scipy-user

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20070821/3f349ea3/attachment.html>