Hugo Pedder

Thanks Josh - solved this problem for me perfectly. Cannot include variables to calculate inits that are not in the jags data in jags.parallel, even though you can with simple jags command

Hi, I am trying to compare model fit statistics between two models, one multivariate and one univariate (see below). I should only be adding one parameter to the model (rho), meaning that the expected number of parameters should only increase by one. Since calculating pD for multivariate models using a manual plug-in method isn't possible (as the saturated model is not clearly defined for the multivariate likelihood), I have been using the pV from JAGS as an approximation to this. However, the pV...

Thanks so much for the help on this Martyn. I've made the ammendments and the code complies fine now. And that's very useful to know about the correlation needing to be -1/M. Should definitely save me quite a bit of trouble in the future. Model is running fine and producing results

Hi, I am trying to perform a meta-analysis in which I need to construct covariance matrices to use in a multivariate analysis, but the covariance matrices are of different dimensions for each study. The covariance matrices are needed to account for the correlation between measurements over time, but each study reports a different number of time points (fups). The problem seems to arise when I take the inverse (or do any further manipulation) with the covariance matrix, as for studies with shorter...