Re: Fitting 2 compartment first-order input PK/PD model.

Date: Tue, 14 Sep 1999 11:19:26 -0700 (PDT) From: ABoeckmann <alison@c255.ucsf.edu> Subject: Re: Fitting 2 compartment first-order input PK/PD model. Thanks to everyone for helpful comments on Dick Wixley's question. Joga's points are especially good ones. A few other remarks: 1) "The output is garbage so I suspect that my input was also." Then why include these? $ESTIMATION MAXEVAL=9999 PRINT=6 SIGDIGITS=6 POSTHOC MSFO=RUN5K.MSF $COVARIANCE Start slowly. Omit the $ESTIMATION record. Look first at the predictions with the initial ests. of theta. If the model is wrong, the Estimation Step will distort the thetas. You won't be able to tell if the predictions are garbage because of the bad thetas or the bad model. This advice is for everyone!!!! 2) As Joga pointed out, there appear to be modelling errors. Probably "K12=AK" should be "K12=AKA" and "K20=AK" is missing. 3) If you do not need elimination from the effect compartment, use ADVAN12. This will give you compt 4 to use as the effect compartment, and is faster than ADVAN7. Alison