A Review and Synthesis of Educational Measurement Procedures for Studying Growth with the Purpose of Specifying the Appropriate Applications for These Procedures. Final Report.

The objective of this study was to review and integrate the various methodologies used in the study of individual growth (especially academic growth). This was accomplished by means of Joreskog's general model for the analysis of covariance structures, i.e., each of the disparate methodologies available from the literature was shown to be a special case of Joreskog's general model. Two general considerations enter into the study of growth and its determinants: (a) making provision for errors of measurement, and (b) constructing a model which relates growth to its determinants in a causally meaningful way. Errors of measurement typically involve questions about the reliability and/or validity of measures, i.e., only indirect measures of the desired variable (construct) are available. Multiple measures of each construct would appear necessary to deal with measurement errors in a quantitative manner. For this purpose the multitrait/multimethod approach devised by Campbell and Fiske (1959) is a useful approach since in principle it allows for correlated errors of measurement. Because the Campbell-Fiske approach does not specify the exact relationships between observed variables and constructs, a factor analytic formulation of their approach was used in order to summarize various approaches to measurement error. The constructs, which represent the growth variable and its determinants, were then interrelated in terms of a linear structural (causal) model. The implications of this model, which itself is a special case of Joreskog's general model, were considered. (Author)