🚀 go-pugleaf

I have a question about Box's M test for homogeniety of  variance-covariance
matrices in a repeated measures MANOVA.

The degrees of freedom I get using SPSS PC+ is very large
 F WITH (3,1726132) DF =           3.34884, P =   .018 (Approx.)

but I can't find any df formulas in any of the references I have handy.


I'm aware (at least my references tell me) that Box's M tends to be overly
sensitive to violations.  I don't really need any heavy theoretical discussions
here.  More than anything else I'm looking for reassurance that SPSS is
computing this correctly and that those degrees of freedom are correct.  If
they aren't, then I've got a different problem.

I include the printout from SPSS below for your information.  There is one
within factor with two levels and one between factor with 2 levels.

Mike Shapiro
Dept. of Communication
Cornell University
MIK@cornella


* * ANALYSIS  OF  VARIANCE -- DESIGN   1 * *

 Cell Means and Standard Deviations
 Variable .. VOTEPOS
      FACTOR           CODE                  Mean  Std. Dev.          N

  BETWEEN                0                   .083       .279         48
  BETWEEN                1                   .044       .208         45
 For entire sample                           .065       .247         93

 - - - - - - - - - -
 Variable .. VOTENEG
      FACTOR           CODE                  Mean  Std. Dev.          N

  BETWEEN                0                   .542       .504         48
  BETWEEN                1                   .156       .367         45
 For entire sample                           .355       .481         93
-------------------------------------------------------------------------------
Page   4                          SPSS/PC+                              2/11/92

* * ANALYSIS  OF  VARIANCE -- DESIGN   1 * *

 Univariate Homogeneity of Variance Tests

 Variable .. VOTEPOS

       Cochrans C(46,2) =                          .64236, P =  .050 (approx.)
       Bartlett-Box F(1,24771) =                  3.77743, P =  .052

 Variable .. VOTENEG

       Cochrans C(46,2) =                          .65366, P =  .034 (approx.)
       Bartlett-Box F(1,24771) =                  4.43012, P =  .035

 - - - - - - - - - -
-------------------------------------------------------------------------------
Page   5                          SPSS/PC+                              2/11/92

* * ANALYSIS  OF  VARIANCE -- DESIGN   1 * *

 Cell Number .. 1

 Determinant of Variance-Covariance matrix =           .01826
 LOG(Determinant) =                                  -4.00312

 - - - - - - - - - -

 Cell Number .. 2

 Determinant of Variance-Covariance matrix =           .00579
 LOG(Determinant) =                                  -5.15247

 - - - - - - - - - -


 Determinant of pooled Variance-Covariance matrix          .01173
 LOG(Determinant) =                                      -4.44575

-------------------------------------------------------------------------------
Page   6                          SPSS/PC+                              2/11/92

* * ANALYSIS  OF  VARIANCE -- DESIGN   1 * *


 Multivariate test for Homogeneity of Dispersion matrices


 Boxs M =                         10.29184
 F WITH (3,1726132) DF =           3.34884, P =   .018 (Approx.)
 Chi-Square with 3 DF =           10.04644, P =   .018 (Approx.)


 - - - - - - - - - -
-------------------------------------------------------------------------------
Page   7                          SPSS/PC+                              2/11/92

* * ANALYSIS  OF  VARIANCE -- DESIGN   1 * *

Tests of Between-Subjects Effects.

 Tests of Significance for T1 using UNIQUE sums of squares
 Source of Variation          SS      DF        MS         F  Sig of F

 WITHIN CELLS              13.22      91       .15
 CONSTANT                   7.90       1      7.90     54.39      .000
 BETWEEN                    2.10       1      2.10     14.43      .000

 - - - - - - - - - -
-------------------------------------------------------------------------------
Page   8                          SPSS/PC+                              2/11/92

* * ANALYSIS  OF  VARIANCE -- DESIGN   1 * *

Tests involving 'ADTYPE' Within-Subject Effect.

 Tests of Significance for T2 using UNIQUE sums of squares
 Source of Variation          SS      DF        MS         F  Sig of F

 WITHIN CELLS              10.18      91       .11
 ADTYPE                     3.77       1      3.77     33.66      .000
 BETWEEN BY ADTYPE          1.40       1      1.40     12.51      .001

 - - - - - - - - - -

      6288 BYTES OF WORKSPACE NEEDED FOR MANOVA EXECUTION.
-------------------------------------------------------------------------------