Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Share this Page URL

11 More Regression Methods > 11.11 ANCOVA - Pg. 609

Chapter 11 More Regression Methods 609 11.11 ANCOVA A common problem is comparing two independent groups of participants in terms of some measure of location while taking into account a covariate. More generally, a common goal is to compare the regression curves corresponding to two or more groups. The Pygmalion data in Section 11.2.1 provides an illustration. If the experimental and control groups are compared using post IQ reasoning scores, the 20% trimmed means are found to be significantly different at the = 0.05 level. A general issue is whether there continues to be a difference when the pretest IQ scores are taken into account. Section 11.2 describes one approach to this problem where the slopes of the regression lines are compared. Before continuing, it is noted that there is a vast literature on the analysis of covariance (ANCOVA) where the goal is to compare groups in terms of some measure of location, usually means, while taking into account some covariate. For an entire book devoted to the subject, see Huitema (1980). For a review of some recent developments, see Rutherford (1992) and Harwell (2003). (For some recent results when using means, see Ceyhan and Goad, 2009.) Obviously all relevant methods cannot be described here. Rather, attention is restricted to situations where comparisons are to be made using some robust measure of location.