multilevel R: how to set covariance between slope and intercepts at zero.

Hi, i know how to do this in Mlwin, but not in R.
i want to reach the most parimonious multilevel model, so i want to check if the covariance between slopes and intercepts is statisticaly significant. To do this i want to define the covariance at zero, and then compare the -2logLikelihood of the models with and the model without covariance between the slope and the model with covariance.

So, actually all I need to know is how to define the coveriance as zero in the lme() function.
so for example i have:
model <- lme(fixed=normexam~standlrt+avslrt, random=~standlrt|school, data=data.tut, method="ML")
so now i need the command for the same model, but with covariance between slope and intercept at zero, but i can't seem to find it...

can anyone help me out?
Last edited:


Phineas Packard
Easy to do if you call winbugs from R using R2Winbugs. Fit can then be assessed in multiple ways. Check out chapter 17 of Gelman and Hill's "Data analysis using regression and multilevel/hierarchical models". Not sure if this is helpful or not though.