comparing two mixed logit models (level vs log indep variable)

[Doris Rivera]

Dear Statalist, I want to compare two multilevel logit models (melogit) with the same dep variable, same indep variables, and same levels. The only difference between them is that the variable of interest is in levels in one model, while it is in logs in the other. Both are significant. Is it possible to compare the log-likekihood, AIC, and BIC for both model, and pick the one with higher log-likelihood, lower AIC and BIC?

An important question is: would it be correct to use pseudo-R2 for each model and compare which one is higher? how do I get the pseudo R2 after the "melogit" model?

Thanks!

[Felix Bittmann]

Yes you can use AIC and BIC for this. For more information regarding R2 see here: http://repec.org/dsug2017/Germany17_Langer.pdf

[Doris Rivera]

Dear Felix, thanks for the suggested pdf. I thought that it would be an easier way to implement some of these Pseudo R2.

[Felix Bittmann]

In my archive I found that I have actually used one ado once for this aim.

findit mlt

'MLT': module to provide multilevel tools

Hope this helps.