Discussion: View Thread

All,

I'm in search of advice as to how to model the following situation (preferably using SAS or Minitab):

Response Variable:  Y = success/failure

Factor 1:  Location Type (1 or 2)

Factor 2:  Treatment (A or B)

Factor 3:  Location (nested within location type) - approximately 10 of each type

Being measured:  many individuals within each location on Treatment A and many individuals within each location on Treatment B.  So factors 1 and 2 are fixed and crossed, while location is a random factor and nested within factor 1.

This seems to me as though it involves the idea of a split-plot design as well as logistic regression; I have experience with both of these separately but never in combining the two.  Can anyone refer me to resources / examples of something similar, again preferably using either SAS or MINITAB (probably SAS)?

Thanks!
-Joe

------------------------------
Joseph Nolan
Associate Professor of Statistics
Director, Burkardt Consulting Center
Northern Kentucky University
Department of Mathematics & Statistics
------------------------------

Hi Joe,

Yes, you've got it right.

It looks to me like you need to do a GLMM (generalized linear mixed model) with location as random and location type and treatment as fixed.  (At least that's my guess based on what you've said).

In SAS you'll want to use proc glimmix.  http://www2.sas.com/proceedings/sugi30/196-30.pdf

If you're familiar with proc mixed, the syntax for glimmix is very similar.

I have no idea if Minitab could do this model.

Karen

------------------------------
Karen Grace-Martin
Senior Statistical Consultant
The Analysis Factor, LLC