## PR#16744: ordering of variance weights (failed in nlme <= 3.1-149) library("nlme") fm3 <- gls(follicles ~ sin(2*pi*Time) + cos(2*pi*Time), Ovary, correlation = corAR1(form = ~ 1 | Mare), weights = ~as.integer(as.character(Mare))) # fixed variance weights stopifnot( identical( getVarCov(fm3, individual = 3), getVarCov(fm3, individual = levels(Ovary$Mare)[3]) ), all.equal( vapply(as.character(1:11), function (id) getVarCov(fm3, individual = id)[1,1], 0, USE.NAMES = FALSE), fm3$sigma^2 * (1:11) ) ) ## lme method had a similar issue for data not ordered by levels(group) ## is.unsorted(Orthodont$Subject) # TRUE fm4 <- lme(distance ~ age, Orthodont, weights = ~as.integer(Subject)) covmats <- getVarCov(fm4, individuals = levels(Orthodont$Subject), type = "conditional") stopifnot( all.equal( vapply(covmats, "[", 0, 1, 1), fm4$sigma^2 * seq_len(nlevels(Orthodont$Subject)), check.attributes = FALSE ) ) ## PR#16806: 1-observation groups in corSpatial fits (failed in nlme <= 3.1-164) data("Phenobarb") pheno <- subset(Phenobarb, !is.na(conc)) stopifnot(sum(getGroups(pheno) == "28") == 1) # Subject 28 has only 1 obs. lme1 <- lme(conc ~ time, data = pheno, random = ~1 | Subject, correlation = corExp(form = ~time | Subject)) condVarCov <- getVarCov(lme1, type = "conditional", individuals = "28") ## gave Error in dimnames(cond.var) <- list(1:nrow(cond.var), 1:ncol(cond.var)) : ## length of 'dimnames' [2] not equal to array extent stopifnot(all.equal(unname(diag(condVarCov[[1]])), lme1$sigma^2)) ## same for gls(): gls1 <- gls(conc ~ time, data = pheno, correlation = corExp(form = ~time | Subject)) margVarCov <- getVarCov(gls1, individual = "28") ## gave Error in rep(1, nrow(S)) : invalid 'times' argument stopifnot(all.equal(diag(margVarCov), gls1$sigma^2))