Estimating hypothesis strengths

Violation of the exhaustiveness postulate of Levine’s (1965) hypothesis model is shown to produce large errors in the estimation of the proportions of problems mediated by the various hypotheses. In two cases, when stat-subject simulations satisfied the postulates, estimates were reasonably close to actual values. One interpretation proposes that negative values of estimates are an indication that the exhaustiveness assumption is violated. An alternative computational method is described.