RE: Output EPS(1) after an $EST step?
Dear Mike,
Even with a simple additive error model the IRES (=DV-IPRED) is not going to
be epsilon, unless the number of observations per subject goes towards
infinity (and the model is correct in all its parts). For an additive +
proportional the corresponding residual (IWRES) is
Y=IPRED+EPS(1)*IPRED+EPS(2)
IRES = IPRED-DV
SIG11=SIGMA(1,1)
SIG12=SIGMA(1,2)
SIG22=SIGMA(2,2)
IWRES = IRES/SQRT(SIG11*IPRED**2+SIG22+SIG12*IPRED)
If SD(IWRES) is approximately 1, the IWRES probably agrees reasonably with
epsilon. If it is lower, there is epsilon shrinkage (also output by NONMEM)
and the agreement between epsilon and the residual is lost.
Best regards,
Mats
Mats Karlsson, PhD
Professor of Pharmacometrics
Dept of Pharmaceutical Biosciences
Faculty of Pharmacy
Uppsala University
Box 591
75124 Uppsala
Phone: +46 18 4714105
Fax + 46 18 4714003
Quoted reply history
From: [email protected] [mailto:[email protected]] On
Behalf Of Dodds, Mike
Sent: 03 February 2012 21:21
To: nmusers
Subject: [NMusers] Output EPS(1) after an $EST step?
All,
Is there a way to recover epsilons after an $EST step? The difference
between DV and IPRED is epsilon for a simple additive error model. However,
with a mixed proportional and additive error model, I can't see a way to
separate the two epsilon contributions to DV-IPRED.
$TABLE EPS(1) after an $EST step provides a column of zeros, and the NONMEM
manual indicates this is the expected behavior.
Warm Regards,
Mike Dodds