Still can't make it work [🇷 for BE/BA]

posted by PharmCat  – Russia, 2020-08-08 03:08 (901 d 20:35 ago) – Posting: # 21838
Views: 11,664

Hi ElMaestro!

❝ That is why you have five elements in theta (T is not replicated, but you are trying to fit both a within and a between variance for it). I am saying we do not work on them both, in stead we replace their sum by a single variance estimate. This is the whole point of the thread here.

I understand your point. And generaly your variance component function is better because is not redundant. From other side if we have some different functions V = ƒ11) and V = ƒ22) it could be any and from any θ if we get equal V (so it could be CSH, FA0(2), manual construct).

❝ If you are looking at a final log likehood around 15.337 then I think we are at least talking about the same model.

I have 15.337299690629834, model is the same ;-) it is good proof: different V construction methods lead to same results.

❝ I guess it must be a little typo in Gurka's paper.

From time to time I find inaccuracies even in good articles and sometime it give strong headake when you try directly reproduce some methods. log(Σ|X|) instead log(|ΣX|) can give hours of frustration.:crying:

❝ Thanks PharmCat, pointing me towards Gurka and insisting on Xi was a great help:ok:.

I am very glad if my thoughts were useful :clap:

Complete thread:

UA Flag
 Admin contact
22,477 posts in 4,708 threads, 1,603 registered users;
23 visitors (0 registered, 23 guests [including 9 identified bots]).
Forum time: 22:43 CET (Europe/Vienna)

We must be careful not to confuse data with the abstractions
we use to analyze them.    William James

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