st: Re: statalist-digest V4 #4330

[Miranda <lanvandy@gmail.com>]

Hello,

I have a statistical question, which may or may not be related using Stata.

I have a categorical outcome variable which has three categories. I
will use the multinominal logistic regression for both the univariate
and multivariate analysis. However, with regarding to the univariate
analysis, their might be such circumstances that certain covariate is
associated with one category of the outcome while not with the other
category. Then how should I make the decision of whether or not to
include this covariate in multivariate analysis? Any other techniques
I could use for the bivariate analysis?

Another question: I am using the respondent driven sampling (RDS)
methods for recruit participants in the study and thus using the
related analytical tool (RDSAT) to do the data analysis. For those of
you who are familiar with this technique, individual or population
weights of the outcome variable are usually exported from the RDSAT
and imported with original data. When doing univariate and/or
multivariate analysis, these weights are usually applied to the
outcome variable. When using Stata, it does have a tab for weights
selection in the logistic regression model. I am not sure whether I
need to choose frequency weights, sampling weights.

Best,
*
*   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/