Page 1 of 1

Covariance matrix for random effects

Posted: Fri Nov 19, 2010 9:22 am
by Corrie
I have a model for repeated measurements with 2 levels, measurement occasions within individuals with a random intercept and several random slope parameters. I’ve noticed that the covariance matrix in MLwiN for the individual-level random effects in the Equations window and the correlations between random effects derived from this are not the same as the correlations between the individual-level residuals produced in MLwiN from the Residuals window. Does anyone know why this is? Is it because of the shrinkage factor applied to the residuals, and if so is there any way of getting residuals from MLwiN which have the same covariance structure as that shown in the Equations window?