Time-varying covariates
The information content of an observation is directly proportional to the derivative of IPRED with respect to the ETA on the parameter evaluated at the EBE for that ETA, asnd inversely proportional to the residual error magnitude.
For each individual and parameter we compute a weighted average of the covariate observations, where the weight is the derivative of IPRED wrt the ETA on the parameter divided by the residual error magnitude at that observation.
This requires the residual error be encoded the 'Uppsala way' using a parameter W for residual error magnitude in the $ERROR/$PRED.
the information content of an observation is directly proportional to the derivative of IPRED wrt the ETA on the parameter evaluated at the eben for that eta, and inversely proportional to the residual error magnitude
EBE: Empiric Bayese Estimate
G is the partial derivative of F wrt eta evaluated at eta bar, which is necessary to calculate the total OBJV for thera, omega, and sigma