Miraculouser oracle [BE/BA News]

posted by d_labes  – Berlin, Germany, 2013-03-12 13:12 (4053 d 10:42 ago) – Posting: # 10190
Views: 7,474

Dear Helmut, dear All!

...

Partial Tests of Model Effects

            Hypothesis  Numer_DF  Denom_DF      F_stat     P_value

----------------------------------------------------------------------

                   int         1        17   1088.59       0

                 Stage         1        17     17.3504     0.000648307

              Sequence         1        17      0.369037   0.551559

        Sequence*Stage         1        17      2.22777    0.153875

Sequence*Stage*Subject        16        17     24.9524     1.04115E-08

          Stage*Period         2        17      0.967076   0.400153

             Treatment         1        17      0.0454252  0.833759


Seems to me like all the F-tests are done with MSE as denominator.
My beasty SAS gives almost the same:
Source                  DF     Type III SS     Mean Square    F Value    Pr > F

treatment                1      0.00208481      0.00208481       0.05    0.8338
period(stage)            2      0.08876896      0.04438448       0.97    0.4002
sequence                 1      0.01436992      0.01436992       0.31    0.5831
stage                    1      0.79630537      0.79630537      17.35    0.0006
sequence*stage           1      0.10224489      0.10224489       2.23    0.1539
subjec(sequen*stage)    16     18.32323177      1.14520199      24.95    <.0001


The question is: Are these tests appropriate?

Like the test for the sequence effect in the classical 2x2x2 crossover I would suppose to test stage and sequence as between-subject effects and also the new introduced interaction sequence*stage with subject(stage*sequence) as denominator. But I'm not quite sure.

The as bogus suspected RANDOM statement in Proc GLM may assist is in getting the 'right' F-tests. But as always when the Almighty miraculous oracle prophesies difficulties are just around the corner (at least for statistical sermons).

Formulating RANDOM stage sequence subject(sequence*stage) / test;
gives me the anticipated denominator
Source                    DF     Type III SS     Mean Square    F Value    Pr > F
treatment                  1        0.002085        0.002085       0.05    0.8338
period(stage)              2        0.088769        0.044384       0.97    0.4002
subjec(sequen*stage)      16       18.323232        1.145202      24.95    <.0001
Error: MS(Error)          17        0.780224        0.045896


Source                    DF     Type III SS     Mean Square    F Value    Pr > F
  sequence                 1        0.014370        0.014370       0.01    0.9122
* stage                    1        0.796305        0.796305       0.70    0.4166
  sequence*stage           1        0.102245        0.102245       0.09    0.7689
Error                     16       18.323232        1.145202
Error: MS(subjec(sequen*stage))
* This test assumes one or more other fixed effects are zero.

But! The F-test for stage is only valid if sequence*stage is assumed zero.

Formulating RANDOM stage sequence sequence*stage subject(sequence*stage) / test;
gives:
Source                    DF     Type III SS     Mean Square    F Value    Pr > F
treatment                  1        0.002085        0.002085       0.05    0.8338
period(stage)              2        0.088769        0.044384       0.97    0.4002
subjec(sequen*stage)      16       18.323232        1.145202      24.95    <.0001
Error: MS(Error)          17        0.780224        0.045896

Source                    DF     Type III SS     Mean Square    F Value    Pr > F
sequence                   1        0.014370        0.014370       0.16    0.7743
Error                 0.7141        0.062612        0.087676
Error: 1.014*MS(sequence*stage) - 0.014*MS(subjec(sequen*stage)) - 22E-16*MS(Error)

Source                    DF     Type III SS     Mean Square    F Value    Pr > F
stage                      1        0.796305        0.796305       7.79    0.2190
Error                      1        0.102245        0.102245
Error: MS(sequence*stage)

Source                    DF     Type III SS     Mean Square    F Value    Pr > F
sequence*stage             1        0.102245        0.102245       0.09    0.7689
Error                     16       18.323232        1.145202
Error: MS(subjec(sequen*stage))

No warning, but ... Rather wild denominators in the F-tests for sequence and stage!
Especially unique for sequence with non-interger df's :surprised:.

Any opinion out there?

BTW: The whole mess is "für die Katz" (strictly for the birds). The confidence intervals are not affected at all.
with sequence*stage 101.454363  (88.445194 ... 116.377016)
without             101.454363  (88.445194 ... 116.377016)

Regards,

Detlew

Complete thread:

UA Flag
Activity
 Admin contact
22,984 posts in 4,822 threads, 1,651 registered users;
50 visitors (0 registered, 50 guests [including 8 identified bots]).
Forum time: 00:55 CEST (Europe/Vienna)

You can’t fix by analysis
what you bungled by design.    Richard J. Light, Judith D. Singer, John B. Willett

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