Date:         Mon, 2 Mar 2009 13:07:05 -0800
Reply-To:     sej.in.md@GMAIL.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         sej.in.md@GMAIL.COM
Organization: http://groups.google.com
Subject:      Re: PROC GLIMMIX: 2 questions: blocks, and covariance matrices
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

On Mar 2, 2:29 pm, rh...@UNMC.EDU (Robin R High) wrote:
> SJ
>
> A couple of important questions:
>
> Why are there so many 0s? a few 1's ? and only a scattered few "counts" >
> 1?  and is this your complete dataset or just a sample?
>
> Keep in mind that an important concept in statistics is the "study of
> variation" .. even for counts (unless your data is more binary in
> reality).
>
> Robin High
> UNMC
>
This is in fact the complete dataset.  What I'm looking at is the
appearance of alternative male reproductive types in termite colonies
after the king is removed: mrs ("male reproductive soldier") and mn
("male neotenic").  [fn is "female neotenic" which show up
unexpectedly from time to time].  So sometimes neither show up,
usually just one or the other, and sometimes both or multiples of one
type.  This is why there are a lot of "0"s.  Even though mrs appears
binary here, it is not naturally binary--I've seen colonies with 2 or
3 of these guys--though these are rare.  This is why I chose a poisson
distribution for the response count.  My goal was to see if 1 of the 2
treatments resulted in more mrs appearing, and the same for the mn.
Sorry for not supplying all the pertinent info at the outset.

-SJ.