[R] Time series count model?
james.lindsey at luc.ac.be
Wed Nov 21 11:17:42 CET 2001
> pauljohn at ukans.edu wrote:
> > I fear I need someone to throw a brick at my head to shake loose
> > the cobwebs. And it might as well be the r friends as anybody!
> > I am trying to counsel a student who has count data (with many
> > 0's and small nuumbers) that is a time series. He was fitting
> > linear models to this data, but the count nature of the data
> > causes me a lot of concern, and I am looking around for a time
> > series approach to a model which allows all the bells and
> > whistles of count models. By chance, I notice that Congdon's
> > WinBUGS examples have an example that has a count model with
> > time series issues, but I fear it might be too great of a leap
> > for us to justify a paradigm switch to Bayesian statistics in
> > order to fit this one model.
> > If you were using R, how would you get a foot hold on this
> > problem?
> I remember that I experimented with some code I hacked according to
> HARVEY, A.C. "Forecasting, structural time series models and the
> Kalman filter"
> who has a chapter about count data and Poisson observations. He suggests
> a time series of Poisson observations, whose mean is chosen by a gamma
> distribution with time-variable paramters (if I recall it correctly).
Essentially this model is implemented in the kalcount function of my
repeated library. Jim
> Maybe that reference might help
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch
More information about the R-help