[S] Generalized Linear Model for Clustered Observations

Gerald Haeubl (Gerald.Haeubl@ualberta.ca)
Thu, 24 Sep 1998 11:32:41 -0600 (MDT)


Hello all,

I wonder if anybody on s-news may be able to offer advice with respect to
the following problem:

My objective is to model the effects of four binary predictors on a set of
dependent variables. Each predictor represents a 2-level factor in the
design of a controlled experiment. Three of these predictors reflect
between-subjects factors, while the fourth corresponds to a within-subject
factor. Some of the dependent variables are frequencies; others are binary
responses. Thus, I would like to fit poisson and logistic regression
models, respectively, using the glm function in S-Plus (4.5 Professional for
Windows). However, since the design contains a 2-level within-subject
factor, the two sets of responses obtained from each individual should not
be treated as independent. As far as I know, glm does not allow for a
repeated measures setup that would take into account the clustered nature of
the data. If the dependent variables were continuous, I would be inclined
to use the lme function and specify subjects as the experimental units. It
seems that what I am looking for is a function that combines the virtues of
glm (allowing non-continuous response variables) and lme (allowing clustered
observations).

Any advice on this would be appreciated tremendously.

Gerald Haeubl

------------------------------------------------------
Gerald Haeubl
Faculty of Business
University of Alberta
Edmonton, Alberta
Canada T6G 2R6
Voice: (403) 492-6886 (office)
Voice: (403) 432-1708 (home office)
Fax: (403) 492-3325
E-mail: Gerald.Haeubl@ualberta.ca
WWW: http://www.bus.ualberta.ca/ghaeubl
------------------------------------------------------

-----------------------------------------------------------------------
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