Mahmoud
★    

Jordan,
2023-03-11 13:21
(13 d 15:25 ago)

Posting: # 23493
Views: 456
 

 average bioequivalence: MLE vs. REML [General Sta­tis­tics]

Dear All
In bioequivalence 2x2 (TR,TR) 3x3 ( TRR,RRT,RTR) and 2x4(TRTR,RTRT)
for the statistical anaylsis based code SAS

PROC MIXED;
CLASSES SEQ SUBJ PER TRT; 1077
MODEL Y = SEQ PER TRT/ DDFM=SATTERTH; 1078
RANDOM TRT/TYPE=FA0(2) SUB=SUBJ G; 1079
REPEATED/GRP=TRT SUB=SUBJ; 1080
ESTIMATE 'T vs. R' TRT 1 -1/CL ALPHA=0.1;
run;
By simulation study I found that

MLE method better when number of fixed effects ≤ 4.
REML method better when number of fixed effects > 4.


Edit: Category and subject line changed; see also this post #1[Helmut]
PharmCat
★    

Russia,
2023-03-14 21:44
(10 d 07:03 ago)

@ Mahmoud
Posting: # 23501
Views: 315
 

 average bioequivalence: MLE vs. REML

❝ MLE method better when number of fixed effects ≤ 4.

❝ REML method better when number of fixed effects > 4.


Hi! Always use REML. Use ML only if you know exactly why.
Mahmoud
★    

Jordan,
2023-03-15 09:35
(9 d 19:11 ago)

@ PharmCat
Posting: # 23503
Views: 310
 

 average bioequivalence: MLE vs. REML

❝ ❝ MLE method better when number of fixed effects ≤ 4.

❝ ❝ REML method better when number of fixed effects > 4.


❝ Hi! Always use REML. Use ML only if you know exactly why.


The greater the number of fixed effects, the greater the difference between REML and ML.
PharmCat
★    

Russia,
2023-03-17 00:51
(8 d 03:55 ago)

@ Mahmoud
Posting: # 23505
Views: 257
 

 average bioequivalence: MLE vs. REML

❝ The greater the number of fixed effects, the greater the difference between REML and ML.


REML is an approach that produces unbiased estimators for some special cases and produces less biased estimates than ML in general. Some people say that REML is unbiased. It is like denominators N and N-1 for variance if you have data from all general population - you can use ML, when you have a small part of population - you should use REML. If you have many independent observations (over 200 for example) then estimates will be nearly same. Some info.
UA Flag
Activity
 Admin contact
22,550 posts in 4,724 threads, 1,606 registered users;
15 visitors (0 registered, 15 guests [including 8 identified bots]).
Forum time: 04:47 CET (Europe/Vienna)

If there is an exception to any rule,
and if it can be proved by observation,
that rule is wrong.    Richard Feynman

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