Two Tukey multiple comparisons and Bonferroni and multivariate approaches are compared for their rates of Type I error and any-pairs power when multisample sphericity was not satisfied and the design was unbalanced. For tests of weighted means and for study conditions investigated, the Bonferroni procedure provides a workable solution. (TJH)

Computer sampling from a multivariate normal spherical population was used to evaluate the type one error rates for a test of sphericity based on the distribution of the determinant of the sample correlation matrix. (Author/LMO)

Multilevel models used to assess and compare school effectiveness, particularly their application to multivariate data, are critiqued. Problems associated with standard errors of measurement and the various components of "school effect" are highlighted in terms of how they could affect incautious application of the models. (TJH)