Hi all,
My problem is the following.
Suppose I have a dataset with observations Y and explanatory variables X1, ...,
Xn, and suppose one of these explanatory variables is geographical area (of
which there are ten, j=1,...,10). For some observations I know the area, but
for others it is unknown and therefore record as NA.
I want to estimate a model of the form Y[i] ~ Poisson( lambda[i] ) with
log(lambda[i]) = constant + \sum_j I[!is.na(area[i])] * I[area[i]==j] * beta[j]
In words: we estimate a constant for all observations and a factor for each
area. If it is unknown what the area is, we only include the constant.
When estimating this model using glm(), the records with is.na(area[i]) are
'deleted' from the dataset, and this I do not want. I had hoped that the model
as described above could be estimated using the function I() (interpret as),
but so far my attempts have not succeeded.
Any help on how to approach this is kindly appreciated.
Kind regards,
Frank van Berkum
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
