Hi -
Is there a way to simultaneously control for temporal autocorrelation and
Poisson-distributed data in R? I have seen similarly threaded topics as
recently as 2009, but the statistical texts suggested in them are well
above my comprehension, and I'm hopeful that more straightforward solutions
have been developed since.
I am trying to analyze population abundance estimates through time --
African lions vs. wild dogs, and lions vs. cheetah, controlling for prey.
My goal is to assess whether there is any potential for suppression of
dogs/cheetahs by lions at the population level. I have 30+ years of annual
data from a single, long-term site where all individuals are recognizable.
>From what I have understood from the literature, my options are:
1) Normalize Poisson population counts with a sqrt() or log()
transformation and run gls() with specified autocorrelation structure.
2) Detrend each time series with a simple linear model and then analyze
residuals with glms.
3) Use geeglm() in geePack to specify both an "association" structure
and
an error structure; however, since I am sampling from a single site, I have
a cluster of 1, and I think this makes the underlying GEE math inviable.
4) And, the unfortunately unavailable for R 2.15.1 but otherwise promising *
Pests* model (http://www.utdallas.edu/~pbrandt/pests/pests.htm) developed
by Brandt, Patrick T. and John T. Williams. 2001. "A Linear Poisson
Autoregressive Model: the Poisson AR(p) Model." Political Analysis 9(2):
164-184.
Thank you all for any advice you can give.
[[alternative HTML version deleted]]