s-news
[Top] [All Lists]

[S] Summary: Dirichlet-multinomial mixture

To: s-news@wubios.wustl.edu
Subject: [S] Summary: Dirichlet-multinomial mixture
From: Melanie Wigg <mbwigg@navigator.math.uwaterloo.ca>
Date: Wed, 24 Feb 1999 11:13:46 -0500 (EST)
Sender: owner-s-news@wubios.wustl.edu
Splus users,

I recently posted the following question:

>I am looking for an Splus function which would estimate the parameters
>of a dirichlet-multinomial mixture (the multicategory analogue of
>the beta-binomial).  I also need this function to handle "counts" that
>are not integers.
> 
>Does anyone know of such an S-plus function?



I received the following responses from ....

Dean Billheimer
----------------

I don't have an S-plus function that directly estimates the
Dirichlet-multinomial parameters.  Several years ago I wrote some Splus
code to perform this feat, and I don't think there was too much involved
(I'm afraid the code fell by the way-side.)  Mosimann (1962, I think)
published a paper in Biometrika that provides expressions for the maximum
likelihood estimators for the parameters.  These can be solved numerically
with nlmin() or nlminb().  As I recall, the only (relatively) tricky part
was to re-parameterize so that the minizer uses only positive values for
the parameter estimates (like exp(x)).  (This may have been before
nlminb() was available.)


Thomas Yee
-----------

I think my VGAM software (http://www.stat.auckland.ac.nz/yee/vgam)
should be able to handle it, provided somebody writes a family function
for it. 


Dave Fournier also replied that he has some software that could do the job.
-------------


Thanks for these responses!
Melanie Wigg
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news

<Prev in Thread] Current Thread [Next in Thread>