A strategy is proposed for dealing with educational research problems, using as an example a study based on data that had already been collected and using that data to discover how differences in school characteristics may result in differences in their students. The research method is based on concepts of analysis of covariance and is concerned with three levels of effects: (1) compositional effects due to initial differences in input variables; (2) contextual effects that have the same effect on all members of a single group; and (3) structural effects that correspond to unequal slopes in regression lines. Three relevant theoretical perspectives are: the small group perspective, interpersonal attraction theories, and the ecological or resource distribution tradition. Interaction effects may provide real clues about actual school differences. The research should begin with a comprehensive structure, and at least two plausible versions of that structure should be proposed with significantly different parameter values. Then the various proposed structures should be compared against the available data and a sensitivity analysis should be conducted to determine whether the data support any one of the structures. (CTM)