COMPUTING COVARIANCE OF PREDICTION ERRORS IN NONMEM

Dear NMUsers, Sometimes, it is desirable to compute an estimate of the covariance of 2 prediction errors in the same individual in order to estimate the variance of the sum of prediction errors in that individual. This estimate has been suggested by Vozeh et al. (1990) to be in the form: where where cov N (PE i ,PE i’ ) is the covariance of the prediction errors i and i’ while E is the expectation of the enclosed _expression_; g and h are partial derivatives of the vector of PK parameters with respect to the ç and å respectively, m is the number of etas in the model etc.) My question is: how can this be coded/outputted from a NONMEM run. Thanks for your help. Jo