Proving Causal Propositions : The Foundations Of Program And Experiment Design

Abstract

Although causal propositions cannot be proven to the point of incorrigibility, they can be disproven (aside from instrument validity problems) or corroborated. Just how one proceeds to such disproof or corroboration depends upon what his interest is in the causes of his dependent variables' values. Testing and qualifying or restricting a specific causal proposition, developing a comprehensive or variance exhaustive linear causal proposition (or multiple regression equation), and mapping or describing the efficacy of a specific set of Treatments imply somewhat,different programs of re- search and experiment designs. Programs and designs for these three interests or strategies are differentiated in terms of a Factor Lattice of all the ex ante relevant variables. The terms of this analysis refer to the regional locations in, density of coverage of, and allocation of replicates to the selected Lattice intersects and the factorial completeness of the design which they constitute, as well as to the type of control exercised over the Factors: production, selection, or stochastic.