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

posted by ElMaestro  – Denmark, 2020-08-07 13:08 (166 d 10:29 ago) – Posting: # 21831
Views: 6,833
(edited by ElMaestro on 2020-08-07 13:25)

Hi PharmCat,

thanks for trying to assist on this. :-) It is kind of you.

» I think Xi is not shuld be full-rank design matrix. It is just part of X for subject i. Moreover all Xi should be the same width (num of columns) for Σ(Xi'V-1Xi). Also with (yi - Xi*β) you get residuals ri (col num of Xi should be equal length β).
» There is not very accurate description in Gurka's paper, but it provide good concept.
» If you check, it should work.
I can't make it work.
In Gurka's equation 3, where we can write the shorthand beta= M1-1*M2: When using the little 3-by-6 matrices for Xi, M1 becomes non-invertible.
I tried to lump 6 subjects together at a time, and now I am getting an estimate of beta but it appears wrong, in the sense it is not corresponding to what I get for the full model.

I wonder if Gurka's equation only works under assumptions not mentioned in the paper or if I am just a very lousy programmer. My money is clearly on the latter :-D
—
Pass or fail!
ElMaestro

Complete thread:

Activity
 Admin contact
21,308 posts in 4,444 threads, 1,489 registered users;
online 4 (0 registered, 4 guests [including 1 identified bots]).
Forum time: Wednesday 22:38 CET (Europe/Vienna)

Power: That which is wielded by the priesthood of
clinical trials, the statisticians, and a stick which they use
to beta their colleagues.    Stephen Senn

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