1 reply to this topic

[vaishnavi ganti]

Hello All Quick question - if we were to perform bootstrap for a PK model with linear and non linear kinetics, the "bands" for Km and Vmax are typically very broad. how would one address this? Also how to get an option to stratify the bootstrap as shown below:

 Thank you

Attached Files

•  bstrap.pdf   126.41KB   585 downloads

Edited by vaishnavi ganti, 06 February 2017 - 08:21 PM.

[mittyright]

Dear Vaishnavi!
As I may understand, the bootstrap procedure implemented in PHX NLME is very simple: each replicate data set is constructed by pooling Nsub random selections of individual data sets with equal probability 1/Nsub on each individual.
So here's no free lunch: you cannot reduce the uncertainty of your model with bootstrap, just amplifying the properties of observed distribution. You may increase the number of replicates to investigate the possible trends

Regarding stratification: you can do that only for categorical covariates

All the best,
Mittyright
PS: please consider to set up the multicore control mode described in PHX NLME user's guide to gain significant performance for bootstrap