Hi Peggy When modelling log-transformed data, it’s useful to add a safeguard for IPRED to prevent taking a log of 0, e.g.:
IPRED = -5 IF (F.GT.0) IPRED = LOG(F) Similarly, when using a proportional (CCV) error model, use the following to prevent division by 0: W = IPRED IF (W.LE.0) W = 0.0001 IRES = DV-IPRED IWRES = IRES/W But as Luann pointed out, for log-transformed data the additive error model (on log scale) is generally used (W=1), which translates to a proportional error on normal scale. Best regards Julia ________________________________________________________________________________ Julia Korell, PhD Pharmacometrician Model Answers Pty Ltd Tel: +61 (0) 7 3221 1315 www.model-a.com.au <http://www.model-a.com.au/> ________________________________________________________________________________ This e-mail, together with any attachments, is confidential and intended for the named recipient(s) only. If you are not the intended recipient or have received this message in error, you are asked to immediately notify the sender and delete this message and any copies of this message from your computer system network and destroy any printed copies of this email. Any form of unauthorised disclosure, modification, distribution, publication or use of this e-mail message is prohibited. > On 25 Aug 2015, at 1:34 pm, luann <[email protected]> wrote: > > Please see my comments inserted below about the values of IPRED,W, etc. for > each type of error model > > I hope the modeling goes well. > > Luann > > On 8/24/2015 9:16 PM, Chiaying Lin wrote: >> Dear NM-Users, >> I'm a NONMEM beginner. The following are my control stream for a monoclonal >> antibody which clears via target mediated disposition. However, when I ran >> NONMEM, I didn't get minimization successful message (in fact, no message at >> all), there are also no Table output and the xx.g77 file contains only empty >> three output files. I will be appriciated if someone help to find mistakes. >> Thanks. >> >> Peggy >> >> --------------------------------------------------------------------------------------------------------------------------------------- >> $PROBLEM **base model** >> $DATA PKDATA.CSV IGNORE=C >> >> $INPUT ID TIME AMT NDV=DROP LNDV=DV EVID MDV CMT BW AGE DOSE >> >> $SUBROUTINE ADVAN6 TRANS1 TOL=3 >> >> $MODEL NCOMP=2 NPAR=6 >> ;COMP=(CENTRAL DEFOBSERVATION DEFDOSE) >> ;COMP=(PERIPH) >> >> $PK >> TVVMAX=THETA(1) >> VMAX=TVVMAX*EXP(ETA(1)) >> KM=THETA(2) >> TVV1=THETA(3) >> V1=TVV1*EXP(ETA(2)) >> TVV2=THETA(4) >> V2=TVV2*EXP(ETA(3)) >> Q=THETA(5)*EXP(ETA(4)) >> S1=V1 >> >> $DES >> C1=A(1)/V1 >> C2=A(2)/V2 >> DADT(1)=-VMAX*C1/(KM+C1)-(Q*C1)+(Q*C2) >> DADT(2)=(Q*C1)-(Q*C2) >> >> $ERROR >> CALLFL=0 >> IPRED=LOG(F) >> IRES=DV-IPRED >> IWRES=IRES/IPRED > > > For CCV error model > IPRED=F > W=IPRED > Y=IPRED+W*ERR(1) > > For LOG error model > IPRED=LOG(F) > W=1 > Y=IPRED+W*ERR(1) > > For additive error model > IPRED=F > W=1 > Y=IPRED+W*ERR(1) > > For additive + CCV error model > IPRED=F > W=SQRT(IPRED**2*SIGMA(1,1) + SIGMA(2,2)) > Y=F + F*EPS(1)+EPS(2) > > The following is true for all of the above models > IRES=DV-IPRED > IWRES=IRES/W > > >> Y=IPRED+ERR(1) >> >> $THETA >> (0.1, 1.49, 5) ;VMAX (mg/hr) >> (1, 50,300) ;KM (μg/mL) >> (1.3, 2.72, 5) ;V1 (L) >> (1.1, 2.43, 5) ;V2 (L) >> (0.001, 0.095, 1) ;Q (L/hr) >> >> $OMEGA .025 ;VMAX >> $OMEGA .032 ;V1 >> $OMEGA .29 ;V2 >> $OMEGA .975 ;Q >> >> $SIGMA >> 0.030 ;ERR(1) >> >> $ESTIMATION METH=1 MAXEVAL=9999 PRINT=5 POSTHOC >> >> $COVARIANCE >> >> $TABLE ID TIME DV IPRED IWRES VMAX KM V1 V2 Q ETA1 ETA2 ETA3 ETA4 BW AGE >> NOPRINT ONEHEADER FILE=1.fit >> $TABLE ID TIME AMT IPRED IWRES NOPRINT ONEHEADER FILE=sdtab1 >> $TABLE ID VMAX KM V1 V2 Q ETA1 ETA2 ETA3 ETA4 NOPRINT ONEHEADER FILE=patab1 >> $TABLE ID BW AGE NOPRINT ONEHEADER FILE=cotab1 >> ---------------------------------------------------------------------------------------------------------------------------------- >> #ID TIME AMT >> >> >> NDV >> >> >> >> >> LNDV >> >> >> >> EVID MDV CMT BW AGE DOSE >> A-P001 0 81.3 . . 1 1 1 81.3 >> 50 81.3 >> A-P001 0.25 >> 4.8 1.568616 >> >> 1 81.3 50 81.3 >> A-P001 1 >> 18 2.890372 >> >> 1 81.3 50 >> 81.3 >> >> >> LNDV=Ln-transformed con. (mcg/ml) > > -- > Luann Phillips | Director, Pharmacometrics > Cognigen Corporation, a wholly owned subsidiary of Simulations Plus, Inc. > 1780 Wehrle Drive, Suite 110 | Buffalo, NY 14221-7000 > Phone: 716.633.3463 x236 | Fax: 716.633.7404 | > [email protected] <mailto:[email protected]>
