Skip to content
Prev 73894 / 398498 Next

Time Series Count Models

We are leveraging too far on speculation, at least from what I can 
see.  PLEASE do read the posting guide! 
"http://www.R-project.org/posting-guide.html".  In particular, try the 
simplest example you can find that illustrates your question, and 
explain your concerns to us in terms of a short series of R commands and 
the resulting output.

	  With counts, especially if there were only a few zeros, I'd start by 
taking logarithms (after replacing 0's by something like 0.5 or by 
adding something like 0.5 to avoid sending 0's to (-Inf)) and use "lme", 
if that seemed appropriate.  Then if I got drastically different answers 
from other software, I would suspect a problem.

	  Other possibilities for count data are the following:

	  * "lmer" library(lme4) [see Douglas Bates. Fitting linear mixed 
models in R. R News, 5(1):27-30, May 2005, www.r-project.org -> 
Newsletter -> "Volume 5/1, May 2005: PDF".

	  * "glmmPQL" in library(MASS).

	  * "glmmML" in library(glmmML)

	  However, I don't know if any of these as the capability now to handle 
short time series like you described.

	  You might also consider the IEKS package by Bjarke Mirner Klein 
(http://www.stat.sdu.dk/publications/monographs/m001/KleinPhdThesis.pdf and
http://genetics.agrsci.dk/~bmk/IEKS.R).

	  spencer graves
Brett Gordon wrote: