Date:         Mon, 22 Nov 2010 10:55:40 -0600
Reply-To:     Warren Schlechte <Warren.Schlechte@TPWD.STATE.TX.US>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Warren Schlechte <Warren.Schlechte@TPWD.STATE.TX.US>
Subject:      Re: Extra-multinomial (Extra-binomial?) variation in SAS
Comments: To: Randall Powers <powers_r@BLS.GOV>
Content-Type: text/plain; charset="us-ascii"

My experience with overdispersion comes from two sources.  Either you
get overdispersion because you have too many zeros, or you have
overdispersion because you have clustering (either temporal or spatial).


To generate count data that has these characteristics, I would think you
would want to build a process model that generates either excess zeros
or clustering, then add an observation model to it.  I don't know of a
procedure that does this, but a simple example shouldn't be too
difficult to code.

Warren Schlechte

-----Original Message-----
From: Randall Powers [mailto:powers_r@BLS.GOV]
Sent: Monday, November 22, 2010 8:26 AM
Subject: Extra-multinomial (Extra-binomial?) variation in SAS

Hey SAS Folks,

I'm wondering if SAS has a function or procedure for generating count
data
subject to extra-multinomial variation (or extra-binomial variation?)??

For example: Have a multinomial (p,n) distribution with p=(p11,
p12,p21,p22) and n=number of units and want to program for n=1000, p=
(.01, .24, .25, .25)

Also, if there are any R people here, is there something in R I could
use
to achieve this as well?

Thanks!