Regression analysis for the comparison of school and home effects

Abstract

This paper addresses two related problems: what are the appropriate measures for comparing effects of blocks or compounds of variables (each block conceived as having a single effect) when the causal ordering is assumed to be known, and how can single measures be created for each such block or compound, so that the comparisons can be made? Several different kinds of measures are shown to be possibly appropriate, through the use of hypothetical experiments.