Intrinsic Variability in Mixed Effects Models
2 views (last 30 days)
Anyone knows how to include intrinsic variability in the form of autocorrelated error in any of the matlab algorithms for mixed effects (nlmefit,nlmefitsa)?
I'll explain better: If I estimate a M.E. model with these algorithms, the solution gives me a beta array (fixed effects) a Psi matrix (covariance of random effects) and an Errorparam array (rmse for additive, proportional or whatever measurement errors I included).
However, the error parameter is general (one for all subgroups), an can be accounted only as measurement error, but it doesn't tell anything about intrinsic variability inside the subgroups (this would mean to have a covariance of the parameters for each subgroup due to intrinsic noise). To account for this, one would need to use stochastic differential equations or maybe include a serial correlation. I've seen that some NLME software do this (Monolix, as far as I know has this feature, or something similar) but I wanted to know if there is something of this already implemented in matlab's toolbox.
Thanks if anyone can help me.