Hello.
I would like to ask you about a procedure to obtain simultaneous exact
confidence intervals for proportions in a multinomial distribution.
The data I am trying to analyze belong to a population where one and
only one of three classes is possible for each person in the
population.
I have a random sample of a few people from that population and I
would like just to give an estimation on the real proportion of people
belonging to each class. I though the best approach was multinomial
CI's, but I've hardly found any reference about exact estimations in
that concrete case.
As my sample size is quite small, I don't feel comfortable with
approaches that use Normal or Chi-square approximations.
Can anyone suggest me a procedure already implemented in R, SPSS or
something similar, and the related bibliographical references (papers,
books...)? Actually, I recently found this option at StatXact 9, but I
don't know which procedure they are using to produce this kind of
exact CI.
I would appreciate any answer. Please, if this is not the right or
best forum regarding my question, let me know.
Thank you very much in advance.
--
vicent
estadis.webs.upv.es
@vginer_upv
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|