skip to main content skip to footer

Unraveling Lord's Paradox: The Appropriate Use of Multiple Regression Analysis in Quasi-Experimental Research

Author(s):
O'Connor, Edward F., Jr.
Publication Year:
1973
Report Number:
RB-73-53
Source:
ETS Research Bulletin
Document Type:
Report
Page Count:
15
Subject/Key Words:
Lord, Frederic M., Multiple Regression Analysis, Pretesting, Statistical Analysis

Abstract

Lord's paradox is a demonstration that the adjustment for pretest differences in preexisting experimental groups depends on assumptions about what the data would have been under circumstances other than those which actually occurred. Different assumptions may lead to conflicting estimates of the treatment effect. In the absence of random assignment, it is often difficult to justify the choice of a particular set of assumptions. The problem is discussed in terms of confounding and specification error, and a procedure is suggested by which a researcher can test her choice of assumptions. The procedure consists in demonstrating that the pretest differences can be explained entirely in terms of a few known variables. It is assumed that all variables are measured without error, although the approach can be modified for fallible variables. (15pp.)

Read More