The remarriage of multiple regression and statistical inference: a promising approach for hypnosis researchers.

Abstract

Abstract A multiple linear regression computer program is described as a general purpose procedure for the testing of hypotheses in hypnosis research. The claim is made, with supporting evidence, that virtually any hypothesis can be tested with the procedure, if the researcher's theory can be placed in the form of a predictive equation where one or more sources of information are used to predict some measure of hypnosis. The hypothesis, stated in the null form, is applied algebraically as a restriction upon the predictive equation, thus yielding a second restricted, predictive equation. The squared multiple correlation coefficients of the two equations are compared in testing the hypothesis. Certain advantages, pedagogical and logical, are claimed, among which is the thought that if science is genuinely concerned with the description, modification, and prediction of behavior, here is a technique which keeps these goals in sight as hypotheses are used to test theories.