I am trying to model counts data from three different sites where the number of
zeros differs significantly from one sight to another and also the counts gets
bigger in site with few number of zeros considerably, meaning that the k's
are different. In my model I am including about 10 covariates to start with, and
I have beed attempting to fit a NEGATIVE BINOMIAL MODEL. The problem is the
function glm.nb neither glm(formula,family=negative.binomial(k),data) work when
I have all covariates in the model. But when we fit only intercept or few
covariates, the algorithm converge. I also fitted the logistic binary (binomial)
for the all three sites using glm with all covariates and there was no problem.
I also fitted the lognormal (counts) model for one of the site using glm and
this worked fine. I have tried to go through the help files in R and Venables
& Ripley 4th Edition but I couldn’t find appropriate solution for my
problem.
Please could you give me idea of how to go about it?
Thanks
BPM
Attached is the part of the out put:
fit <- glm.nb
PFD~SEX+RES+TRAV+TRAVH+HEIGHT+WEIG+MUA+TEM+FEVV+VOM+HEA....,data=data)
Error: NA/NaN/Inf in foreign function call (arg 1)
In addition: Warning message:
iteration limit reached in: theta.ml(Y, mu, n, limit = control$maxit, trace =
control$trace >
---------------------------------
Want to chat instantly with your online friends? Get the FREE Yahoo!Messenger
[[alternative HTML version deleted]]
