LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (February 2008, week 3)Back to main SAS-L pageJoin or leave SAS-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Wed, 20 Feb 2008 11:17:47 -0500
Reply-To:     Jerry Davis <jwd@UGA.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jerry Davis <jwd@UGA.EDU>
Subject:      Re: mixed-effects loglinear model
In-Reply-To:  <200802200936.m1K5flI0019125@malibu.cc.uga.edu>
Content-Type: text/plain; charset="US-ASCII"

Mei-Ling Bai wrote:

Details are a bit sketchy, but maybe my comments will be useful to you.

> I want to investigate the relationship between variable1, variable2 and > variable3. But this data was based on the observation of wildlife for > several years, thus many individuals were observed repeatedly for > different years. I think I may apply a loglinear model with SAS proc > NLMIXED, using indvidual as the random effect.

The procedure GLIMMIX has the ability to fit log-linear models and is easier to use than NLMIXED. It also has some nice options for plots, inverse link means etc. GLIMMIX is available for download from SAS Institute for SAS version 9.

> > Is this approach proper? Shoupd I use individual as random intercept or > random coefficients? Could someone suggest a reference for NLMIXED code of > loglinear models?

If you are using GLIMMIX these two statements are equivalent:

RANDOM individual; RANDOM INTERCEPT / SUBJECT=individual;

The second statement is preferable because it scales up better due to the way GLIMMIX sets up the model. If you have messy data or lots of subjects, using the second statement may yield convergence, while using the first statement may not.

Jerry

-- Jerry Davis Experimental Statistics UGA, CAES, Griffin Campus


Back to: Top of message | Previous page | Main SAS-L page