Use of hypotheses for analysis of variance models: challenging the current practice

Abstract

In social science research, hypotheses about group means are commonly tested using analysis of variance. While deemed to be formulated as specifically as possible to test social science theory, they are often defined in general terms. In this article we use two studies to explore the current practice concerning group mean hypotheses. The first study consists of a content analysis of published articles where the reconstructed reality of hypotheses use is explored. The second study is a qualitative interview study with researchers, adding information about daily practice. We argue that, at present, hypotheses are not used to their utmost potential and that progress can be made by using informative hypotheses instead of the current non-informative hypotheses. Informative hypotheses capitalize on knowledge that researchers already possess and enable them to focus in their proceeding projects. The substantive focus of our work is the case of applied psychology.