Date: Fri, 8 Nov 2002 13:32:26 -0500 joe.mendez.fernandez@utoronto.ca "SPSSX(r) Discussion" Joseph Mendez Re: Conditional logit regression in SPSS using COXREG To: Rob Penfold <000301c28743\$97dd5a60\$e9b65f41@penfoldr5enqac> TEXT/PLAIN; charset=US-ASCII

What is your research question formulated as a 1,0 outcome? Maybe I didn't catch it in the question, my error.

All you need to trick COXREG to do a CLR is the time variable which is 1 for cases and 2 for everyone else. This is just tricking SPSS to calculate the "soon-ness" of being a case, it doesn't mean anything mathematically. The outcome variable is the same as a regular logistic regression, 0 and 1. The advantage is it lets you use partial likelihood with strata for the pairs..

The SPSS trick is described in one of the Appendixes to Kleinbaums book Logistic Regression which is at the UofT Bookstore (the Yellow new edition) -Joseph

On Fri, 8 Nov 2002, Rob Penfold wrote:

> I am having trouble figuring out how to structure my data file in order to > perform a conditional logit regression using the COXREG feature in SPSS. I > have a set of discrete locations and I know the location decisions of each > store in my analysis. I have 150 stores and 120 discrete locations. The > location characteristics (regressors) differ for different stores faced with > choosing to locate there. Does this mean that I need a 120x150 matrix of > choices in addtion to the choice characteristics? And, if the choice > characteristics differ for different choosers, how do I build this further > dimension into a two dimensional table? > > When I finally do this, can I test the IIA assumption in SPSS? > > Many thanks, > > Rob Penfold > PhD candidate > Department of Geography > University of Toronto >

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