genZcor fails if using unstructured correlation structure with only two observations per cluster

[jwilliman]

Hello

If there are only two observations per cluster, and an unsturctured correlation structured is used, genZcor and consequently geeglm fail with a relatively unhelpful message.

gendat <- function() {
  id <- gl(5, 4, 20)
  visit <- rep(1:4, 5)
  y <- rnorm(id)
  dat <- data.frame(y, id, visit)[c(-2,-9),]
}
set.seed(88)
dat <- gendat()
dat2 <- subset(dat, visit %in% 1:2)

zcor <- geepack::genZcor(clusz = table(dat$id), waves = dat$visit, corstrv=4)
zcor <- geepack::genZcor(clusz = table(dat2$id), waves = dat2$visit, corstrv=4)
#> Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]): contrasts can be applied only to factors with 2 or more levels

^{Created on 2023-03-27 with reprex v2.0.2}

This appears to be caused by the line ans <- model.matrix(~z - 1) in genZcor, where z is a factor variable with only one level (1:2).

Can this be made to work, or a helpful error message added explaining the problem?

Thanks very much for your work on this package :)