Hi everyone,
I would like to run an lme analysis on a response variable that has unequal
variability among samples. This variability is quantified by a standard
error estimate for each measure of the response variable. Prior to the lme
analysis, the response variable is calculated in a series of generalized
linear models (one for each measurement of the response variable). I
realize that the variability among estimates of the response violates an
important assumption of lme analysis, and I am attempting to find the
correct method of adjusting for these differences in the lme analysis. An
idea that a colleague suggested to me is that I may be able to incorporate
the standard error from each of these estimates in the lme analysis.
I have developed the following potential solution to the problem - does this
seem to be correct?
First, the ususal command lines that I use:
lme462$PopnYearNetDay<-paste(as.character(lme462$Popn),
as.character(lme462$Year), as.character(lme462$NetDay),sep=":")
YfitM <- lme(M~NetDay+NetType+LakStr+Cond+DCoarsef,
random = pdBlocked(list(pdIdent(~Popn-1), pdIdent(~as.factor(Year)-1),
pdIdent(~PopnYearNetDay-1))), data=lme462)
Then, the following questionable command:
fmYfitM <- update(YfitM, weights = inerrorM)
Here, inerrorM would be the inverse of standard errors from the estimates of
the response variable in the Generalized linear model.
Thank you very much for any help that you can give me with this.
Sincerely,
Cindy Rejwan
Cindy Rejwan
Ph.D. Candidate
Department of Biological Sciences
University of Calgary
Current Address:
Fisheries Centre
University of British Columbia
2204 Main Mall
Vancouver, British Columbia
V6T 1Z4
|