WinNonlin is lack of precision or? [Software]

posted by yicaoting  – NanKing, China, 2011-11-06 15:20 (5343 d 14:44 ago) – Posting: # 7619
Views: 13,519

Dear warm-hearted ElMaestro,
Thank you for your quick response.

❝ The fitting engines must stop at some point when the objective function cannot be improved much further. The software internally keeps track of the improvement at each iteration and sets some criteria for stopping. It can for example be when the Log likelihood (or SS) cannot be improved by more than 0.0001 arbitrary units or something like that.

❝ ... There are many different ways to optimise; some prefer the brutally efficient Nelder-Mead algo, others the softcore Newtonian-based meffuds like Marquardt-Levenberg. Entire books are written about this.


But you as everyone know (as WNL's usr guide pointed, Phoenix WinNonlin 6.0 guide Page 338), to convert
90% CI of LSMean(R-T) to 90% CI of GeoMean Ratio(%) (T/R), we only need a simple calculation like
CI_Lower = 100 · exp(lower)
CI_Upper = 100 · exp(upper)

No iteration is needed, and no optimization algorithm is needed.

Or we can conclude that WNL uses a strange (and of course low-precision) way to calc 90% CI of GeoMean Ratio(%) (T/R), but not the way pointed in the User Guide.

❝ By the way, with R 2.10.1 I get:

format(qt(0.95,22), digits=20)

[1] "1.717144374380243"


Thank you for your kind help. My so called Second Found is clear now: key point is WNL's Tinv()'s low-precision, of course I have to accept it.

Complete thread:

UA Flag
Activity
 Admin contact
23,655 posts in 4,993 threads, 1,571 registered users;
125 visitors (0 registered, 125 guests [including 17 identified bots]).
Forum time: 07:05 CEST (Europe/Vienna)

Scientists often have a naïve faith that
if only they could discover enough facts about a problem,
these facts would somehow arrange themselves
in a compelling and true solution.    Theodosius Dobzhansky

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5