From: Eleveld, DJ d.j.eleveld@anest.umcg.nl
Subject: Am I interpreting THETA(1) incorrectly?
Date: Thu, March 24, 2005 9:04 am
Hello NONMEM users,
I am a NONMEM beginner and am trying some prelimiary tests and have found an
(at least to me) confusing result. Possibly a more experienced NONMEM user
can explain the situation.
We have parameters that are log-normally distributed using V1=THETA(1)*EXP(ETA(1))
in the $PK section. From this notation it seems to me that the estimated value for
THETA(1) should be the geometric mean value of the individual V1 values. However
when I do POSTHOC to get the individual V1 values, the geometric mean of these values
is different than the estimated typical V1 value (i.e. THETA(1)).
Can someone explain why this occurs? as the estimated geometric mean of the
parameter population values?
As an aside the data I am fitting comes from a monte-carlo simulation and the geometric-mean
of the POSTHOC individual parameter values is closer to the 'real' values than the THETA values.
Thank you,
Doug
Am I interpreting THETA(1) incorrectly?
From: "Liping Zhang" ZHANG_LIPING@lilly.com
Subject: [NMusers] Re: Am I interpreting THETA(1) incorrectly?
Date: Thu, March 24, 2005 10:28 am
Doug, there is nothing mystery here. Just like if you random sample 10 normally distributed
numbers with mean equal to 0, the mean of these ten numbers are not likely to be 0. Theta
would be equal to the geometric mean of the posthoc individual parameter if posthoc
individual parameters are log normal distributed and truly represented the distribution
(theoretically you need infinite sample size).
As for why the geometric mean of posthoc individual parameters are closer to the real
values of theta, this could be caused by mere chance. You could do experiment: simulation
different set of data and then compare the two.
Liping Zhang, PhD
Global PK/PD Modeling and Trial Simulation
ELi Lilly & Company
Work: 317-277-8687 Fax: 317-433-6661
DC 0734
Email: zhang_liping@lilly.com
_______________________________________________________
F