I am analysing a discrete choice experiment. I joined the project at a later stage, so I was not involved in the design of the experiment per se.
The core of the DCE was to see how different ways of framing information on energy efficiency affect the uptake of more efficient products. For this reason, three versions have been run: a control version plus two treatments which display energy efficiency in alternative ways. In the control group, the attribute energy efficiency can take on just 2 values (Energy Star yes, or Energy Star no). On the other hand, in the treatment groups energy efficiency can have 6 levels (1-6 based on monetary expenditures for electricity).
I want to run a mixed logit model pooling the three groups together. This is usually done by creating interactions between the attribute variable (EE) and treatment dummies (T1 and T2)
Code:
EE_T1 = EE * T1 EE_T2 = EE * T2
Code:
mixlogit choicer Price Stars OptOut, group(cluster_id) rand(Capacity Brand_n EE EE_T1 EE_T2) id(ID) cluster(ID) nrep(1000)
Any suggestion will be greatly appreciated.
Thank you all in advance.
Stefano
0 Response to Problem with attribute having different dimensions across treatment groups for mixed logit model estimation
Post a Comment