Dear all
Last week, I posted a question and some of you spent their valuable time to
respond. Thank you (once again). This time, after doing a lot of homework, I
return with a second question on the issue.
I was advised to use poLCA for a Latent Class Analysis with covariates. The
software works well, although I feel a bit uncomfortable with the fact that you
need to run many times the algorithm to reach the global maxima (so how do we
know that we replicated the analysis enough times???). This software solved one
of my problems, but now I need a software to be able to compute probability of
laten group membersips using ordinal "response" variables and at the
same time fit covariates.
I have seven ordinal variables of Political Trust (scale from 0-10 where 0 means
no trust and 10 means full trust). Around 50,000 persons completed the
questionnaire. I would like to compute the probability of membership for 4 (it
seems that this is about the right number) of clusters. Then, I will need to use
a small number of covariates to predict membership. Unfortunately, poLCA cannot
compute probability of latent group membership for ordinal categorical
"response" variables. Can aynone suggest another R package (or any
other free package)?
Or maybe is there an R package to use my ordinal "response" variables
to compute membership probabilities and then (as a second independent analysis)
use another software to compute regression coefficients where the membership
probabilities will be coninuous(???) dependent variables?
Thnk you for the help
Jason
Dr. Iasonas Lamprianou
Department of Social and Political Sciences
University of Cyprus
>
[[alternative HTML version deleted]]
