Dear Listers:
I am not sure whether you have receive my previous email regarding -clogit-
in Stata.
If yes, please forgive my second try.
I am estimate a multiple choice model.
In this model, totally I have around 30000 individuals. (In Stata, that
means, I have around 30000 groups)
The number of choice in individual's choice set is around 50.
Couple choice specific characteristics are included. (around 10 variables,
say denoted as Zj variable)
However, I also have 4 individual specific variables (say, denoted as Xi
variable)
As suggested by Greene and Stata Manual, in order to identify the
coefficient for those Xi variables, I need to interact those Xi(individual
characteristics) with choice dummies.
Since I have around 50 choices, therefore, by this interaction, totoally, I
need to add around 200 new created variables.
My model is as following:
Left hand side variable: Choice (this is a binary variable, constructed
following Stata's description)
Right hand side variables: (1) Zj( choice specific charactersitics)
(2) Xi*Choice Dummy
Therefore, totally, my right hand side variable contains more than 200
variables.
My first try of estimation used around 10 hours for obtaining restuls.
The second model(similiar to the above, but different dataset) used 12 hours
for convergence.
I am using Window XP, and CPU is P4 2.66G Hz with 1G memory.
I am just wondering whether anyone has this kind of experience that Stata
needs many many long time for convergence when using -clogit-
Or, does anyone know that I can use another method or setup in order to save
more time?
Or, in fact, -clogit- takes long long long long time for convergence when
the number of right hand side varialbe is more than, say, 200?
Thank you very much
JT
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
