Coefficient constraints using asclogit

[Chris Cronin]

Hello,

I am using asclogit to estimate a model with both alternative-specific and case-specific variables. I would like to impose constraints on the case-specific variable coefficients so that they are the same across specific groups of alternatives. I believe I need to use the "constraints()" option, but do not understand exactly how to use the option. Below, I provide a short example of my problem.

Assume n individuals are selecting among 50 automobiles and they have the option to not purchase (i.e., there are k=51 alternatives). I have an alternative-specific independent variable in "mileage" and a case-specific variable in "male." If I run:

asclogit choice mileage, case(id) alternatives(automobile) casevars(male)

then state will estimate 1 mileage coefficient, 50 male coefficients, and 50 constant coefficients. I would like to hold the effect of "male" fixed across automobile options so that being male affects only the likelihood of purchasing any car, rather than the likelihood of selecting a particular car. Stata would then only return 1 male coefficient. I would also like to know how to hold the constant coefficient fixed across alternatives.

Thank you.

[Arne Risa Hole]

I would not use the "contraints()" option in this case. Here is an alternative solution - this assumes that the "automobile" variable equals 51 for the "no purchase" option:

gen car_asc = automobile != 51
gen car_male = car_asc * male
asclogit choice mileage car_asc car_male, case(id) alternatives(automobile) noconst

Arne

[Chris Cronin]

Thank you, Arne. This is solution is much easier that what I had in mind.