The missing data library allows multiple imputation. The object
produced by the MI is a list. I want to run a simulation that uses MI
on data sets that have missing data that are MAR. My initial attempts
to have each MI data set extracted and analyzed using miSubscript were
unsuccessful because the extracted object was a list.
To illustrate, suppose that I created an imputed data object called
dat.imp. The object has 5 imputed data sets that I want to analyze and
record some of the results.
mylist <- letters[1:5]
for (i in 1:5){
x <- miSubscript(dat.imp, i)
assign(x = mylist[i], value = cbind(x[rep(1:nrow(x), rep(4,
nrow(x))),1:2], time = rep(0:3, nrow(x)), y =
as.vector(t(data.matrix(x[,3:6])))))
z <- rbind(z,summary(glmmPQL(y ~ time, random = ~ 1|id, fam =
binomial, data = mylist[i]))$tTable)
}
I use Splus 6.1, Win2K.
Respectfully,
Frank R. Lawrence
|