A Note on Testing for Multivariate Effect Sizes

This note provides the investigator with an overall test procedure for testing the significance of multivariate effect sizes, assuming a regression model for the vector of effects, using Wald's statistic. Given a regression model of the form δ = Xβ, the step down Finite Intersection Test (FIT) procedure is used to test that all effect sizes are zero. Finally, an overall study effect size is proposed. Large sample tests are developed that involve the multivariate chi-square distribution.