| Abstract: | Contends that both the interpretation of an effect size and the actual estimation of a coefficient of determination are partially theory-dependent. Two theoretical models for the variables cases are considered. In a variety of circumstances where the square of the correlation is used, the required assumptions are not tenable. In the alternate model, the absolute value of the correlation provides a coefficient of determination. The correlation coefficient is recommended for use as an effect-size indicator, because evaluating effect size in terms of variance accounted for may lead to interpretations that grossly underestimate the magnitude of a relation. (25 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved) |
