Bioequivalence and Bioavailability Forum • random vs. fixed

random vs. fixed [Software]

posted by Helmut – Vienna, Austria, 2011-10-07 23:35 (4955 d 14:32 ago) – Posting: # 7447
Views: 6,825

Hi Yicaoting & ElMaestro!

❝ ❝ For dataset 3, results from WNL:

❝ ❝ LSM_R: 82.5594 (WNL) vs 82.5594 (SAS)

❝ ❝ LSM_T: 79.6926 (WNL) vs 79.2074 (SAS)

❝ Hrmmmmmmmfff...

Yessir.

❝ However, there is -at least in theory- an alternative when one value is missing for a period in one (or more) subject(s) and that is to try a maximum likelihood approach where you specify subject as random in the mixed model and trt+seq+per all fixed. When I do that in R, I actually can reproduce your values from WNL (but I do not have WNL on my machine so cannot play around). It could thus be that WNL actually uses a mixed model to obtain the estimates?

Right guess. Phoenix/WinNonlin’s default in BE is:

fixed:  Sequence+Formulation+Period

Random: Subject(Sequence)

Therefore we get REML estimates. If we delete the random effect and specify the model as ‘all fixed’
Sequence+Formulation+Period+Subject(Sequence)
WNL will spit out exactly SAS’ results (LSM_T, SE, Diff, and CI) for dataset 3.

❝ Someone, go read the manual?!

Wasn’t necessary. ;-)

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

Least Square Means (LSM) for incomplete data yicaoting 2011-10-07 18:27 [Software]
- Least Square Means (LSM) for incomplete data ElMaestro 2011-10-07 21:00
  - random vs. fixedHelmut 2011-10-07 21:35
    - random vs. fixed ElMaestro 2011-10-07 23:18
      - random vs. fixed Helmut 2011-10-08 14:48
        
        random vs. fixed ElMaestro 2011-10-08 15:26
        
        random vs. fixed Helmut 2011-10-08 16:02