search for: reml2

...likelihood value does not agree with this definition. As an example consider the following code: library(nlme) data(Glucose) model <- lme(conc~Time, data=Glucose, random=~Time|Subject, na.action=na.omit) LL.ml <- logLik(model,REML=F) #=688.588 LL.reml <- logLik(model,REML=T) #=692.654 LL.reml2 <- as.numeric(0.5*log(det(model$varFix)) + LL.ml) #=694.489 Here LL.reml2 is computed according to the above formula and it does not agree with the REML value as computed in lme. Does anybody know why is this so and how I may evaluate log.REML at a set of parameter values using log.ML? Regards...