Statistical Sciences 2035 Study Guide - Quiz Guide: Repeated Measures Design, Mixed Model, Multivariate Testing In Marketing
16 Oct 2018
School
Department
Professor
Document Summary
Box test: non-significance = equal covariance assumption is satisfied. Covariances are unstandardized correlations: how 1 variance relates to another variance. Partially tests the homogeneity of variance assumption. If satisfied, the correlations between the levels of the ws factor will also bear close resemblance in both groups. Not robust if the dependent variables within the various populations do not all have a normal distribution. In practice: dependent variables often deviate from normality box test is often unreliable. In practice: if we have equal sample sizes per group we wouldn"t have to look at box test, because we can do anova anyways. Study output (glm repeated measures: multivariate & univariate) Only has ws main effects and wsxbs effects. First: look at interaction here: significant main effects are meaningless. Check sphericity with mauchly"s test: when the ws factor has only 2 levels sphericity is always met (that"s the case here) If sphericity is violated, always use epsilon correction!
