RES and WRES

From: "Leonid Gibiansky" leonidg@metrumrg.com Subject: [NMusers] RES and WRES Date: Fri, May 27, 2005 3:47 pm My understanding was that WRES (weighted residuals) are obtained from RES (residuals) by some transformation (below is the description from the NONMEM html help) that is roughly equivalent to the division by a variance. To my surprise, I found out that for some observations RES and WRES have different signs (METHOD=1, INTERACTION, combined error model implemented in log-transformed variables Y=F + SQRT(THETA(1)+THETA(2)/F**2)*ERR(1) ). So the question is when it happens and what does it mean. Thanks ! Leonid WRES The weighted residuals for an individual are formed by transform- ing the individual's residuals so that under the population model, assuming the true values of the population parameters are given by the estimates of those parameters, all weighted residu- als have unit variance and are uncorrelated. As with the predic- tion and residual, the weights are also computed at eta = 0.

From: "Andrew Hooker" Andrew.Hooker@farmbio.uu.se Subject: RE: [NMusers] RES and WRES Date: Sat, May 28, 2005 5:33 am Hi Leonid, The weighted residuals are calculated by dividing the vector of each individual's residuals (res_i) by the square root of the matrix of covariances of that individual's data conditional on the population model: WRES_i = RES_i / SQRT(COV(data_i | F_pop)) This means that for each WRES calculated we include covariances between data points of an individual. If the correlations between some of these data points are negative then the resulting WRES could also be negative, while the RES could be positive. -Andy Andrew Hooker, Ph.D. Div. of Pharmacokinetics and Drug Therapy Dept. of Pharmaceutical Biosciences Faculty of Pharmacy Uppsala University Box 591 751 24 Uppsala, Sweden Tel: +46 18 471 4304 http://www.farmbio.uu.se/research.php?avd=5 _____________________________________________________