search for: infjack

...hts=w) #were each observation has its own id. The problem with this solution is that gee does NOT allow weights (probably something that would be fairly simple to modify); thus, the gee option doesn't really work. For those who are interested, one work around is to use the Thomas Lumley's infjack.glm function in the glmerrs.R file at http://faculty.washington.edu/tlumley/glmerrs.R and the code is all in the WEAVE package at http://faculty.washington.edu/tlumley/weave.html In the example, notice that the SE estimates are much closer to the SE values in the original model without weights. &...