[NMusers] Number of bootstrap runs needed for confidence intervals

Dear nmusers,
 
I have a question about bootstrap techniques.
 
I have developed a 2-compartment PK model and I wanted to evaluate the=
 model with the bootstrap technique. After 200 successful bootstrap runs=
 obtained with Wings for Nonmem I calculated the parameter means of the 200=
 runs.
Am I correct that at least 1000 bootstrap runs are needed to get correct=
 confidence intervals?
(But due to long run times and access to only 2 CPUs 1000 runs can’t be=
 performed unfortunately.)
If I calculate confidence intervals from the 200 runs, how well is the=
 quality of them? Can I use them at all?
 
Also I’m interested in your opinion about the following method:
A prerequisite for this nethod has to be that the 200 bootstrap runs and by=
 this the parameter values are a representing distribution of the true=
 parameter population. Is it possible to perform a bootstrap from the=
 bootstrap results? Is it possible to get 1000 new bootstrap parameter=
 results by bootstrapping from the 200 old bootstrap values. And finally=
 could I calculate from these 1000 parameter values the confidence=
 intervals and compare them with my original model parameter values.
 
From my point of view this procedure can only be possible if we have the=
 representing distribution of the 200 runs. But how could we be sure, that=
 this is true?
 
Any help or opinion is appreciated!
 
Thanks in advance, Kati



________________________________
Katharina Kuester
Pharmacist
PhD student
Freie Universitaet Berlin
Institute of Pharmacy
Dept. Clinical Pharmacy
Kelchstr. 31
D-12169 Berlin

Phone: +49-30-838 506 27
Fax: +49-30-838 507 11
e-mail: kkuester_at_zedat.fu-berlin.de



Received on Mon Apr 21 2008 - 10:05:28 EDT