bear v2.0.1 (upwards) cont'd [🇷 for BE/BA]

posted by ElMaestro  – Denmark, 2008-10-28 10:49 (6076 d 19:17 ago) – Posting: # 2588
Views: 33,938
(edited on 2008-10-28 13:04)

Dear Dr Lee,

I am sorry if my signals have been wobbly.
I think at some point I read a thread on the mailing list for R http://tolstoy.newcastle.edu.au/R/ about the simple linear model, that it -in R's implementation- might be problematic or some other negative thing (will give the specific link if I find it again, promise) in case of imbalance. I could easily have misunderstood it, no doubt.
Due to this I suggested lme, which has some potential advantages:
  1. It will give the correct result.
  2. It should work for the more complex designs, too, not just the standard 2,2,2-BE design.
Since then I tried to challenge lm and glm in R with various stupid datasets and could not find that it is problematic (that is, I generally obtain the exact same results with lm/glm as with lme regardless of the imbalance in the dataset). On this basis I am inclined to think now that for a 2,2,2-BE design lm is as good as glm which is as good as lme. And of course, some might say lme is overkill for such simple datasets where in the strictest terms we do not need to work explicitly with mixed models.

Finally, in some *AHEM COUGH*iers I have seen for 2,2,2-BE studies, the protocol has explicitly stated that proc glm would be used if the dataset was balanced, and that proc mixed would be used in case of imbalance. I am not a statistician so I cannot qualify this in any way, I only relay it as an observation.

Best regards
EM.

Complete thread:

UA Flag
Activity
 Admin contact
23,424 posts in 4,927 threads, 1,673 registered users;
29 visitors (0 registered, 29 guests [including 24 identified bots]).
Forum time: 07:06 CEST (Europe/Vienna)

Truth and clarity are complementary.    Niels Bohr

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