Comparing mixlogit results across equations

[jamal mamkhezri]

Dear statalist users,
I run two separate mixlogit (Arne Hole commend) models and I want to compare corresponding coefficients. This is what I'd do if I use conditional logit (clogit):

1- run model 1, store estimation (X)
2- run model 2, store estimation (Y)
3- test [X_choice]variable=[Y_choice]variable

However, Stata does not find equations after running mixlogit. Why is that? How can I compare coefficients after running two separate mixlogit models? The same problem occurs if I estimate my models in WTP-space (mixlogitwtp). I appreciate any help.

Thanks,
Jamal

[jamal mamkhezri]

For future reference:
Suppose you have a treatment group and a control group. You would like to know if your variable (say X) that differentiates the two groups from each other is binary (1- treatment, 0- control). The way that I came around the first problem was that I included an interaction term of all of my main-effect variables with variable X and then tested to see if those interacted variables are statistically significantly different from zero.

The one caveat with this approach is: if you assume a lognormal distribution for your price variable (or whatever that goes to the denominator of mWTP equation) and include an X-price interaction term, then calculating WTP for your main-effect variables will become tricky. This will be an issue in both mixlogit and mixlogitwtp commends. I have not found a way around this YET. Any help/thought would be highly appreciated.