Back to the Top
The following message was posted to: PharmPK
Is it possible to perform a weighted curve fit with Berkeley Madonna? That is, can the data be weighted by some function representing reciprocal variance. I note the method for weighting difference data sets when performing curve fitting but this does not seem to allow weighting individual data points. Log transforming the data appears to allow weighting by 1/val^2. Am I missing something?
Thank you, David Bourne
Back to the Top
In the Berkeley Madonna version 8.3.11 that I frequently use for PKPD modelling, there is no provision for weighted Least Square analysis. In PK data analysis, it is recommended that we use Ordinary Least Square (OLS method) analysis without applying weights only if the concentration data is used after log transformation. Usual practise for PK model fitting is ILS or Iterated Least Square analysis where 1/(predicted Y square) is used as weight in successive iterations.
Cheers.
Suresh M.Solapure
Back to the Top
The following message was posted to: PharmPK
Dear Suresh and all,
I would suggest that the best thing to do with weights is not to
mess with Mother Nature. First, determine the assay SD (not CV%!) at about 5
sample points covering the range of the assay. Divide these samples into at
least 5 (more is better) aliquots. Measure each aliquot and determine the
assay mean and SD at each point. Next, fit the relationship between assay
conc and its SD with a polynomial of up to 3rd order. Now you have a good
estimate of the SD for any sample that goes through your assay system.
Now, use a population modeling software package such as the NPAG
software in the MM-USCPACK collection, and estimate the remaining
environmental noise in the system either by estimating Gamma, a
multiplicative factor, of Lambda, an additive factor. In this way you now
know how much of the noise in the system is due to the assay and how much to
the environment. Useful info. And again, you are not INVENTING weights. You
are not messing with Mother Nature, which is a big no-no.
You might look at:
Jelliffe RW, Schumitzky A, Van Guilder M, Liu M, Hu L, Maire P, Gomis P,
Barbaut X, and Tahani B: Individualizing Drug Dosage Regimens: Roles of
Population Pharmacokinetic and Dynamic Models, Bayesian Fitting, and
Adaptive Control. Therapeutic Drug Monitoring, 15: 380-393, 1993.
Bustad A, Terziivanov D, Leary R, Port R, Schumitzky A, and Jelliffe R:
Parametric and Nonparametric Population Methods: Their Comparative
Performance in Analysing a Clinical Data Set and Two Monte Carlo Simulation
Studies. Clin. Pharmacokinet., 45: 365-383, 2006.
All the best,
Roger Jelliffe
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Weighted curve fitting with Berkeley Madonna" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)