Dear all,

I need to perform outlying identification for analysis of longitudinal data (highly unbalanced in the sense that each subject has different number of measurements) in WinBUGS.

My model is a linear mixed model with both random intercept and random slope. The model is implemented in WinBUGS. I want to use CPO (Conditional Predictive Ordinate), but to be able to do that I need firstly to calculate PPO (Posterior Predictive Ordinate). I have know that it involves defining the density of multivariate normal in WinBUGS. Given the fact that, I have unbalanced data, does any one could help me to define the variance-covariance matrix \Sigma_i for subject i in WinBUGS?

Thank you for any idea or any source of reference document.