Hello Statalist Users,
I am using -fmlogit- to fit a fractional multinomial logit model for
panel data. The data has the following form:
ID year share1 share2 share3 ID-specified vars (X)
case-specified var (Price1 Price2 Price3)
1 2000 0.1 0.5 0.4 *
*
2 2000 0.2 0.7 0.1 *
*
3 2000 0.3 0.2 0.5 *
*
1 2001 0.2 0.1 0 7 *
*
2 2001 0.5 0.4 0.1 *
*
3 2001 0.1 0.6 0.3 *
*
It works great if not considering the "case-specific" independent
variable "Price1-3".
Now I want to add "Price" to the model. So the equations become:
share_of_case1 =
exp(Xb(1)+price1*c1)/(exp(Xb(1)+price1*c1)+exp(Xb(2)+price2*c2)+exp(Xb(3)+price3*c3))
share_of_case2 =
exp(Xb(2)+price2*c2)/(exp(Xb(1)+price1*c1)+exp(Xb(2)+price2*c2)+exp(Xb(3)+price3*c3))
share_of_case3 =
exp(Xb(3)+price3*c3)/(exp(Xb(1)+price1*c1)+exp(xb(2)+price2*c2)+exp(Xb(3)+price3*c3))
-mlogit- can fit constrained models (with subsets of coefficients
constrained to be zero) by specifying the constraint() option.
So I can fit the above model by including price1-3 in all equations
and set c2=c3=0 for eq1, c1=c3=0 for eq2, and c1=c2=0 for eq3.
But it seems -fmlogit- does not support the constraint() option.
So is there a way that I can fit the above model using -fmlogit- or
any other command?
Your help would really be appreciated.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/