suzette at sdac.harvard.edu
Wed Aug 13 03:23:25 CEST 2003
I am trying to use nlme to model a large data set
of pharmacokinetic concentrations with sparse data per cluster.
The base model is a one compartment model with first order absorption.
The program runs but the PNLS step never converges. The parameter
estimates are reasonable. The plot of the individual level residuals is
pretty reasonable and the plot of the residuals from the population model
is ok but not great.
My questions are:
1) I know the PNLS step starts and updates the fixed parameters
and random effects. Step 2 is the LME step and it updates the interpatient
variance. Does the LME step include both the EM algorithm and a Newton-
Raphson ms step?
2) I set Maxiter=500 interations and I get convergence on what
I think is the LME step but PNLS step never converges. Any suggestions
on what I might do to achieve convergence would be very much appreciated.
Suzette Blanchard, Ph.D.
Frontier Science Foundation
1244 Boylston St. Suite 303
Chestnut Hill, MA 02467
Email: suzette at sdac.harvard.edu
Phone: (617) 632-2007
Fax: (617) 632-2001
More information about the R-help