Have you considered Fahrmeir and Tutz (2001) Multivariate Statistical
Modelling Based on Generalized Linear Models, 2nd ed. (Springer)? This
book focuses primarily on multinomial responses, but their results apply
for bernoulli variates, which they also consider in some examples. Ch.
6 discusses "Fixed Parameter Models for Time Series and Longitudinal
Data", with especially categorical response variables. Ch. 8 considers
"State Space and Hidden Markov Models".
Can you tell us more about your application(s)? I'm currently using
library(mvtnorm) to provide a latent variable model driving a bernoulli
process. I have not yet estimated parameters for this kind of model,
but it look like it should be fairly straightforward.
hope this helps.
spencer graves
Abderrahim Oulhaj wrote:
Dear All,
I am looking for any statistical method or any related reference for
modelling the sum of n dependent and non-identically distributed
bernoulli variables (i.e. the probability of succes is not constant) for
relatively small n (i.e. no central limit theroem can be used)....
Many thanks,
Abderrahim
Abderrahim Oulhaj, Phd in Statistics
Oxford University
Department of Pharmacology
Mansfield Road
Oxford OX1 3QT
Tel: +44 1865 224098
Fax: +44 1865 224099
Email: abderrahim.oulhaj@pharmacology.oxford.ac.uk
<mailto:abderrahim.oulhaj@pharmacology.oxford.ac.uk>
|