Proportional and Additive Error model with transformed data

1 messages 1 people Latest: Jun 09, 2008
Title: Paul R I am trying Mats' suggestion for the additive + proportional error model for some data I am trying to model. (Parent drug and two sequential metabolites). The model converges successfully when I use the proportional model (eg., Y1 and Y2. However, when I add the additive component (Y3), the run stops immediately with the message: 0PROGRAM TERMINATED BY OBJ ERROR IN OBJ2 WITH INDIVIDUAL 6 ID=0.60000000E+01 CONDITIONAL CONTRIBUTION OR IT'S DERIVATIVE IS INFINITE MESSAGE ISSUED FROM ESTIMATION STEP AT INITIAL OBJ. FUNCTION EVALUATION What is my error? I thought I was preventing a divide by zero with my FF=1 unless .GT. 0. I look forward to your wisdom. The $ERROR section of Control stream follows. Thanks in advance. Paul $ERROR (OBS ONLY) SW1 = 0 SW2 = 0 SW3 = 0 IF(CMT.EQ.1) SW1 = 1 IF(CMT.EQ.2) SW2 = 1 IF(CMT.EQ.3) SW3 = 1 FF=1 IPRED=LOG(FF) W1 = THETA(9) W2 = THETA(10) W3 = THETA(11) W4 = THETA(12) IF (F.GT.0) IPRED=LOG(F) IF (F.GT.0) FF=F IRES = DV-IPRED IWRES1 = IRES/W1 IWRES2 = IRES/W2 IWRES3 = IRES/W3 IWRES = SW1*IWRES1 + SW2*IWRES2 + SW3*IWRES3 Y1 = LOG(FF)+SQRT(W1**2) * EPS(1) Y2 = LOG(FF)+SQRT(W2**2)*EPS(1) Y3 = LOG(FF)+SQRT(W3**2+ W4**2/FF**2)*EPS(1) Y=Y1*SW1 + Y2*SW2 + Y3*SW3 -- Paul R. Hutson, Pharm.D. Associate Professor UW School of Pharmacy 777 Highland Avenue Madison WI 53705-2222 Tel 608.263.2496 Fax 608.265.5421 Pager 608.265.7000, p7856