MANOVA and Levene's Test Significant Values

[autumnleaf]

Hi

I've conducted a MANOVA with 11 dependent variables.

All the other assumptions have been met, except the Levene's test. Levene's
test of Equality of Error Variances is significant for 3 of the dependent variables.

Does this mean, the MANOVA SPSS results will not be meaningful?

Will it be better to report t-test results for the 11 variables?

Or is it recommended to report t-tests for those 3 variables with significant values on Levene's test?

Or would it be better to re-do the manova with the remaining 8 variables and exclude the 3 with significant Levene's test values?

I have not found any book explaining clearly what should be done if Levene's test is significant.

Please HELP!

Thank you,

Simm

[terzi]

The assumptions in MANOVA are regarding the covariance matrices. You should test for equality in the covariance matrices across groups. SPSS calculates the M-Box statistic to test that.

Still, there are many issues related to the M-Box statistic. If you don't "pass" the test (that is, you reject the null hypothesis) one criterion, subjective but useful, is to examine the log of the determinants from all the covariance matrices for each group and also for the error matrix. If those numbers are similar there shouldn't be much trouble.