Hello, I'm recently using lclogit2 to work out a discrete choice model of 2 alternatives with 120k observations. When I use the "lclogitml2" after "lclogitml", I meet an "initial values not feasible" problem.
I've tried several methods including:
1) run the example code in lclogit2. Example works, which proves the completeness of the function;
2) remove most of the membership variables for a simpler model, won't work;
3) downsample most of the observations, won't work;
Since these methods won't work, I had a view of the source code of lclogitml2.ado and suspect the error may happen here:
Can you give me some suggestions on where the error might exist? Thank you.
Code:
lclogit2 choice, rand(fare time) group(id) id(id) nclasses(4) membership(match_rate demand miles speed #here are another 20 variables#) seed(0) #successfully get the output of lclogit2 matrix start = e(b) lclogitml2 choice, rand(fare time) group(id) id(id) nclasses(4) membership(match_rate demand miles speed #here are another 20 variables#) seed(0) from(start) initial values not feasible r(1400);
1) run the example code in lclogit2. Example works, which proves the completeness of the function;
2) remove most of the membership variables for a simpler model, won't work;
3) downsample most of the observations, won't work;
Since these methods won't work, I had a view of the source code of lclogitml2.ado and suspect the error may happen here:
Code:
ml model gf0 lclogitml2_gf0() `Class' `Share' `Fix' if `touse', /// missing maximize nopreserve init(`from') search(`search') `constr' /// `options' wald(0)
Comment