AW: Weighting observations

Hi Leonid, hi Rob, Thank you very much for your suggestions. I have implemented Leonid's fixed error as follows: ARV = THETA(1) PRV = THETA(2) PRV1 = PRV/4 IF (CMT.EQ.1) IPRED = (A(1)-IIVRE-VIN/1000) IF (CMT.EQ.1) IRES = DV - IPRED IF (CMT.EQ.1) W = SQRT(ARV*ARV+PRVV*PRVV*IPRED*IPRED) IF (CMT.EQ.1) DEL = 0 IF (CMT.EQ.1.AND.W.EQ.0) DEL = 1 IF (CMT.EQ.1) IWRES = IRES/(W+DEL) IF (CMT.EQ.1) Y = IPRED+EPS(1)*W IF (CMT.EQ.2) IPRED = A(2)/A(1) IF (CMT.EQ.2) IRES = DV - IPRED IF (CMT.EQ.2) W = SQRT(IPRED*IPRED*PRV*PRV) IF (CMT.EQ.2) DEL = 0 IF (CMT.EQ.2.AND.W.EQ.0) DEL = 1 IF (CMT.EQ.2) IWRES = IRES/(W+DEL) IF (CMT.EQ.2) Y = IPRED+EPS(1)*W IF (CMT.EQ.3) IPRED = A(3)/A(1) IF (CMT.EQ.3) IRES = DV - IPRED IF (CMT.EQ.3) W = SQRT(IPRED*IPRED*PRV*PRV) IF (CMT.EQ.3) DEL = 0 IF (CMT.EQ.3.AND.W.EQ.0) DEL = 1 IF (CMT.EQ.3) IWRES = IRES/(W+DEL) IF (CMT.EQ.3) Y = IPRED+EPS(1)*W IF (CMT.EQ.4) IPRED = A(4)/A(1) IF (CMT.EQ.4) IRES = DV - IPRED IF (CMT.EQ.4) W = SQRT(IPRED*IPRED*PRV*PRV) IF (CMT.EQ.4) DEL = 0 IF (CMT.EQ.4.AND.W.EQ.0) DEL = 1 IF (CMT.EQ.4) IWRES = IRES/(W+DEL) IF (CMT.EQ.4) Y = IPRED+EPS(1)*W IF (CMT.EQ.5) IPRED = A(5)/A(1) IF (CMT.EQ.5) IRES = DV - IPRED IF (CMT.EQ.5) W = SQRT(IPRED*IPRED*PRV*PRV) IF (CMT.EQ.5) DEL = 0 IF (CMT.EQ.5.AND.W.EQ.0) DEL = 1 IF (CMT.EQ.5) IWRES = IRES/(W+DEL) IF (CMT.EQ.5) Y = IPRED+EPS(1)*W $SIGMA (1 FIX) I had to add an additive error on W [CMT=1] since it can take negative values. It considerably improved my fit IPRED vs. DV concerning the first compartment (which is also my compartment of highest interest) without penalizing too much the rest of my observations, although my OFV increased). This solution has several limitations: - the loss of the proportional residual variability on each compartment, which was the case in my original model - the arbitrariness of the fixed factor I haven’t implemented the other three solutions proposed by Leonid since it appears to me to be difficult to handle this problem sequentially. My next step will be to duplicate/triplicate/...? the observations from the first compartment in my data set and in the model file keeping the proportional residual variability on each compartment. @Rob I am not modelling the effect of a drug but the ultrafiltration (intra-abdominal volume balance) achieved with the continuous peritoneal dialysis. I am observing the drained volume and its glucose, urea, creatinine and protein concentration given the instilled volume with known glucose concentration and plasmatic concentration of glucose, urea and creatinine. The flows are described my mono-exponential or bi-exponential equations. Since the lecturers of this forum are mostly pharmacometricians, I have just tried to give a telling example. And I fully agree with you that the first step should be to improve my structural model before tweaking my error model. However, my current model does already account for the hierarchical interaction between compartments (from a physiological point of view) and seems to work with individual data. Any advice? Many thanks in advance. Robert ______________________________________________ Robert, Kalicki, Dr. med. OA Neph INSELSPITAL, Universitätsspital Bern DURN Klinik und Poliklinik für Nephrologie, Hypertonie und Klinische Pharmakologie Freiburgstrasse 15, KiKl, G, 524 3010 Bern Telefon: +41 (0)31 632 31 44 Mobile: +41 (0)79 239 98 79 Fax: +41 (0)31 632 97 34 E-Mail: [email protected] [email protected] www.insel.ch -----Ursprüngliche Nachricht----- Von: Leonid Gibiansky [mailto:[email protected]] Gesendet: Freitag, 21. März 2014 15:47 An: Robert Kalicki; [email protected] Betreff: Re: [NMusers] Weighting observations Hi Robert, I have not seen it done, but you can try to enforce the weighting by playing with the error model. For example, if you have 2 DVs, PK and PD, run the model without any weighting and with the proportional error for both variables: $ERROR SDPK=THETA(1) SDPD=THETA(2) IF(PK) Y=DVPK+SDPK*EPS(1) IF(PD) Y=DVPD+SDPD*EPS(2) $SIGMA 1 FIXED 1 FIXED and got THETA(1)=0.1 and THETA(2)=0.2 you may try SDPK=THETA(1) SDPD=4*THETA(1) effectively forcing the model to downgrade the PD observations. Yet another possibility is to fit that main variable first, fix the parameters related to that variable (population or even population and individual), and then add data for the second variable. Or may be you can estimate the model with only the first variable, and then fix the residual error on the estimated value: this would also force the model to maintain the quality of the first-variable fit. More mild control is to fit that main variable first and then add priors to the model, more informative for the parameters that describe the first variable and uninformative for the second one. If you implement any of these ideas, please let the group know how it worked (or not). This discussion is somewhat similar to the question whether to perform sequential or simultaneous fit of the PK and PD data, so you may look on the references, posters, discussion topics related to this question for additional ideas. Thanks Leonid -------------------------------------- Leonid Gibiansky, Ph.D. President, QuantPharm LLC web: www.quantpharm.com e-mail: LGibiansky at quantpharm.com tel: (301) 767 5566