BootstrapFit()

[nsk64]

Hi everyone,

Performing a bootstrap n=500 with nonmem often leads to some unsuccessful runs. The terminated runs are not always useful and could be discarded if their number is "small". I do not have in mind an commonly accepted rule.
What about nlmixr? and how to deal with that?

Best regards
Nicolas

[mattfidler]

Looking at the code in nlmixr2extra, bootstraps that are unsuccessful return NA for the parameter estimates, so they are ignored. I'm unsure what a good number is, but I typically use 10% myself.