Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

How to replicate SMRT logit results in SPSS?

I was wondering if it is possible to replicate the results I get from the aggregate logit in SMRT in SPSS? I am trying to use the multinomial logistic regression in SPSS, but the results are very different so I was wondering if somebody has some clues on what settings need to be used or perhaps the syntax they use themselves?

I know that SMRT zero centers the utilities but I'm hoping just to get the same std errors and t values so that I understand exactly what SMRT does in the background when I estimate the model.
asked May 26, 2012 by anonymous
retagged Sep 13, 2012 by Walter Williams

2 Answers

0 votes
Perhaps there is a difference in the missing value treatments.
answered May 29, 2012 by anonymous
0 votes
I am unfamiliar with the exact analysis in SMRT, but to analyze the DC data, you should use a conditional logit regression model. In SPSS (I am using version 18) this is possible by doing a cox-regression, not the standard multinomial logistic regression (since this is not a conditional model). Perhaps if you try this model, your parameters will match exactly.

the cox-regression is exactly the same as the STATA 'clogit' function.
answered May 29, 2012 by Sander (360 points)