Abstract
Analysis of covariance (ANCOVA) is frequently recommended to correct for covariate imbalance and to control conditional size of hypothesis tests in randomized designs. Such control, however, depends on the accuracy of the ANCOVA model used to make the adjustments. Examination of residuals together with knowledge of the subject matter are important for obtaining an accurate model. Stratified assignment procedures lessen the dependence of proper control of conditional size on the specific model used in the adjustment procedure.

This publication has 5 references indexed in Scilit: