Prev: need assistance on the SPSS code understanding
Next: output correlation matrix from PROC REG to a dataset
From: Jay Weedon on 27 Jul 2010 17:11
Is this possible? I can't figure out how to introduce the extra scale
parameter that would be needed. I know it can be done with GENMOD.

Thanks, Jay.
From: Ryan on 27 Jul 2010 17:45
On Jul 27, 5:11 pm, Jay Weedon <jwee...(a)earthlink.net> wrote:
> Is this possible? I can't figure out how to introduce the extra scale
> parameter that would be needed. I know it can be done with GENMOD.
>
> Thanks, Jay.

Jay,

To allow for overdispersion for count data in the GLIMMIX procedure,
you could fit a negative binomial model ("dist=negbin") or a
generalized poisson model (http://support.sas.com/documentation/cdl/en/
statug/63033/HTML/default/viewer.htm#/documentation/cdl/en/statug/
63033/HTML/default/statug_glimmix_sect026.htm).

Is this what you're after?

Ryan
From: Jay Weedon on 28 Jul 2010 11:16
On Tue, 27 Jul 2010 14:45:53 -0700 (PDT), Ryan
<ryan.andrew.black(a)gmail.com> wrote:

>On Jul 27, 5:11�pm, Jay Weedon <jwee...(a)earthlink.net> wrote:
>> Is this possible? I can't figure out how to introduce the extra scale
>> parameter that would be needed. I know it can be done with GENMOD.
>>
>> Thanks, Jay.
>
>Jay,
>
>To allow for overdispersion for count data in the GLIMMIX procedure,
>you could fit a negative binomial model ("dist=negbin") or a
>generalized poisson model (http://support.sas.com/documentation/cdl/en/
>statug/63033/HTML/default/viewer.htm#/documentation/cdl/en/statug/
>63033/HTML/default/statug_glimmix_sect026.htm).
>
>Is this what you're after?
>
>Ryan

Thanks Ryan, that's very helpful. I did look over the GLIMMIX doc
examples, but somehow missed that one. I remain a little puzzled about
a couple of features of it though, guess I'll follow up with tech
support:

1. The docs don't seem to refer to the automatic variable _phi_ that
appears in the example (perhaps that omission is what I was missing
before I posted);
2. The covariance test for scale works only if the model also contains
random f/x.

Jay.
 | 
Pages: 1
Prev: need assistance on the SPSS code understanding
Next: output correlation matrix from PROC REG to a dataset