MISUSE OF ANALYSIS OF COVARIANCE WHEN TREATMENT EFFECT AND COVARIATE ARE CONFOUNDED

EXAMINES THE CURRENT PRACTICE OF USING ANALYSIS OF COVARIANCE (AC) WHEN THE TREATMENT IS CORRELATED WITH THE COVARIATE, TO ADJUST TREATMENT MEANS FOR DIFFERENCES IN THE COVARIATE MEANS. EXAMPLES SHOWING THAT FALSE SIGNIFICANT RESULTS CAN BE OBTAINED THROUGH CONFOUNDING THE VARIANCE OF THE COVARIATE WITH THAT OF THE VARIATE, OR THAT EVEN WHEN VARIATE MEANS HAVE PERFECT LINEAR CORRELATION WITH THE COVARIATE MEANS, ADJUSTMENT MAY LEAVE A SIGNIFICANT TREATMENT EFFECT. THESE DIFFICULTIES RESULT FROM A FAILURE OF THE ASSUMPTION OF HOMOGENEITY OF BETWEEN-AND WITHIN-GROUP REGRESSION. CONSIDERATION OF THE FUNDAMENTAL MODEL FOR AC SHOWS THAT THIS ASSUMPTION SHOULD BE EXPECTED TO FAIL IF TREATMENT AND COVARIATE ARE CORRELATED. (23 REF.) (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1968 American Psychological Association.