search

jhhughes256 / optinterval

Raw R code for manuscript: Hughes JH, Upton RN, Reuter SE, Phelps MA, Foster DJR. Optimising time samples for determining AUC of pharmacokinetic data using non-compartmental analysis.
0 stars 0 forks source link

Can a threshold on standard error be used to improve fits? #19

Open jhhughes256 opened 6 years ago

jhhughes256 commented 6 years ago

Issue Description

Standard errors derived from hessian matrices of optim function output could be used to assess confidence in that output.

Hessian matrices could be included in final output of both broad and narrow study designs. This would then allow assessment of standard errors for the best and worst fits.

jhhughes256 commented 6 years ago

Progress

With the use of the new optim.interv.dt function it seems that even bad standard errors can result in good results (with no non-bioequivalent results using AUC and Cmax being observed).

The question now is, will refitting of bad standard errors improve the algorithm? Or is it unnecessary additional computation?