[Limdep Nlogit List] Variable choice sets

[Mikołaj Czajkowski]

On 2010-06-18 17:40, Yijing Lu wrote:
> Dear Mikolaj Czajkowski,
>
> Thank you for your reply. In my model, there are attributes varying across
> the choices and invariant variables across choices. Therefore, if I use
> ;Rhs=, I will have to use the ;Rh2=. However, the difficulty for me is how
> to specify the 'list' in the utility functions U(list)=, since the choice
> sets are different for different people.
>

I fail to see the problem (and this probably means I didn't fully 
understand it) but why don't you simply specify a utility function in 
which you define all the attributes at once (e.g. 
U(*)=b0+b1*x1+b2*x2+b3*x3+....). If in some choice sets some of the 
attributes are missing they will simply take a value of 0 and hence will 
not influence the estimates of their parameters. E.g. if in one of the 
alternatives x2 and x3 were missing the utility function will become: 
U=b0+b1*x1+b2*0+b3*0+b4*x4+....=b0+b1*x1+b4*x4+....
This is providing that you have generic (not alternative-specific) 
parameter estimates but you can easily work around it.

Best regards,

     Mikolaj Czajkowski