Bioequivalence and Bioavailability Forum • Potvin

Potvin – all effects fixed - SAS [Two-Stage / GS Designs]

posted by d_labes – Berlin, Germany, 2012-10-29 10:57 (4620 d 00:24 ago) – Posting: # 9479
Views: 11,567

Dear Helmut!

Here the sasophylistic answer (to what question ever):

Used code:

title "Long form analysis after stage 2";

Proc GLM data=Potvin2_long;

  class treatment period subject sequence stage;



  model lnPK = treatment period(stage) sequence stage subject(stage*sequence)

               /ss3 CLParm alpha=0.0588; * =2*0.0294;

  * random serves only for correct tests for sequence stage;

  random stage sequence subject(sequence*stage) /test;

  estimate 'T-R' treatment -1 1;

  lsmeans treatment / stderr cl pdiff alpha=0.0588;

run; quit;

As we know meanwhile the random statement doesn't change the "all fixed effects" solution of Proc GLM.

Output (only interested parts):

 ...

                    Least Squares Means



                                                             H0:LSMean1=

                                  Standard    H0:LSMEAN=0      LSMean2

 treatment     lnPK LSMEAN           Error       Pr > |t|       Pr > |t|



 R              1.13343123      0.04840026         <.0001         0.8338

 T              1.14787011      0.04840026         <.0001





 treatment     lnPK LSMEAN     94.12% Confidence Limits



 R                1.133431        1.035392     1.231470

 T                1.147870        1.049831     1.245909





         Least Squares Means for Effect treatment



               Difference

                  Between     94.12% Confidence Limits

   i    j           Means     for LSMean(i)-LSMean(j)



   1    2       -0.014439       -0.151665     0.122787





         94.12% (1-2*alpha) CIs - backtransformed



                   point         lower         upper



              101.454363     88.445194    116.377016

Attention! Must had change the sign (swap upper, lower) in the log domain since R-T is calculated due to lexical order of treatments coded as R and T.

Remark for the nitpickers and silly-o-meter inventors:
Within SAS you must use the subject effect nested within sequence and stage, even if the subjects are coded unique. Otherwise you don't obtain a type III ANOVA and no LSMeans.

—
Regards,

Detlew

Complete thread:

Potvin – all effects fixed (PHX/WNL vs. SAS, R) Helmut 2012-10-26 19:35
- Potvin – all effects fixed (PHX/WNL vs. SAS, R) ElMaestro 2012-10-26 20:14
  - Potvin – all effects fixed (PHX/WNL vs. SAS, R) Helmut 2012-10-26 20:25
    - Potvin – all effects fixed (PHX/WNL vs. SAS, R) ElMaestro 2012-10-26 22:05
- Potvin – all effects fixed - SASd_labes 2012-10-29 09:57
  - Potvin – all effects fixed (PHX/WNL resolved) Helmut 2012-11-06 01:43
    - Potvin – all effects fixed (PHX/WNL resolved) ElMaestro 2012-11-06 11:48
      - Potvin – all effects fixed (PHX/WNL resolved) Helmut 2012-11-06 13:48
        
        Mysterious ways ElMaestro 2012-11-06 15:27
        
        Mysterious ways Helmut 2012-11-06 17:06