The Application of Analysis of Covariance in Social Science Research

T HIS paper treats the problem of matching experimental groups used in sociological research. It proposes the substitution of mathematical matching techniques, for case matching, where the data permit treatment by regression and analysis of variance as these two combine to form analysis of covariance. It attempts to clarify the function of analysis of covariance: (i) by showing how, in the classical before-after experimental design, the significance of the difference between final expected and final observed scores is the basic element in analysis of covariance; (2) by showing how regression techniques and analysis of variance combine to form the mathematical manipulations of covariance; and (3) by comparing the meaning of variance, analysis of variance, and analysis of covariance. It presents a pattern, which the student may follow in setting up a problem in covariance, designed to permit the computation and interpretation of a covariance quotient, without his being thoroughly versed in all the algebra of covariance. Finally, it illustrates how covariance may be used to test differences which cannot be tested by simple ratio techniques.