I have several demographic variables with which I want to explain the
ordered choice of individuals within a survey in an ordered choice (probit
or logit, this is not important) framework. Standard ordered choice
estimations of course just give me aggregate/average parameter estimates.
For my task it would however be useful to estimate or extract "hypothetical"
individual-level parameter estimates (betas) for a certain independent
variable and each individual in the survey.
I have experimented with hierarchical Bayes algorithms provided by the
bayesm and ChoiceModelR. Correct me if I am wrong but I think these
techniques also demand that individuals to appear several times within a
survey (thus it should be a panel) and are confronted with different choice
situations, so that one can estimate the influence of certain attributes on
the individuals choices. Anyway ChoiceModelR and bayesm just provide
multinomial choice models while I am seeking for an ordinal probit.
My data however doesn't have any panel structure. I was also experimenting
with Bayesian inference in example by the MCMCoprobit function in the
MCMCpack package, but this function just simulates betas. I can't however,
as far as I know, attribute them to certain individuals in the survey, which
would be good. I would be very glad if somebody could give me a hint,
sometimes already a catchword is helpful to google the correct solution!
Thanks and best regards,
AK
P.S.: the last thing I tried was Compound Hierarchical Ordered Probit
(CHOPIT) because with that I am able to calculate individual cut-off points
which maybe allow be to calculate individual betas. but i didn't try it
exetnsively yet.
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
