Parallel groups in bear - CIs [🇷 for BE/BA]

posted by ElMaestro  – Denmark, 2010-04-22 23:47 (5088 d 15:27 ago) – Posting: # 5188
Views: 55,977

Dear d_labes,

❝ Seems to me as there is something wrong with the random part.


OK, then here's a potential theory:
if we do lme(lnAUC0t ~ drug, random=~1|subj, data=TotalData, method="REML" )
on a dataset where we have parallel non-replicated groups (one observation per subject) then R might be trying to actually fit the error sigma2 on the diagonal of V (R matrix), and for ZGZt possibly ending up with a subject sigma2 on the diagonal. Since there is only one observation per subject, I guess this would amount to a V with error-sigma2 plus subject-sigma2 on the diagonal.
Now error-sigma2 plus subject-sigma2 can be estimated, but the individual components cannot (would explain anomalies, at least). We would effectively only be working with a single sigma2.
In this case, I would be inclined to imagine that the error-sigma2 plus subject-sigma2 would equal the residual error from an anova on lm(lnAUC0t~drug) or something like that.

Best regards
EM.

Pass or fail!
ElMaestro

Complete thread:

UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,639 registered users;
71 visitors (0 registered, 71 guests [including 7 identified bots]).
Forum time: 14:14 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

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