Hi all,

The Mixed Logit model allows to estimate taste heterogeneity by assigning some attributes (toward which the researcher suspects that preferences may be heterogeneous) as random vars.

however, it is not clear whether it makes sense to assign interaction var. between respondent's observable and alternative's attributes.. Even to the extent that the interaction variable is binary, (a multiplicity of 2 binary variables) I'm sure if there is any sense to assign it as a random variable.

Any hint?

Anat