Fractional Factorials in Psychology: An: Examination of the Basis for the Holistic Approach

Abstract

Statistical methods and experimental designs are used by the researcher to collect and analyze data. One of the most commonly used treatment designs is the analysis of variance. Simon (1976) has examined data from fourteen years of experiments published in the Human Factors journal that were analyzed using this approach. One of his conclusions is that interactions of three or more variables are negligible. To take advantage of this Simon (1973, 1977) has proposed using fractional factorials in psychological research. This investigation examined Simon's analysis and conclusions and questions the use of fractional factorials without proper theoretical justification. An empirical investigation of fractional factorials using computer simulation techniques examined the effect of confounding when the assumption that high order interactions are negligible is violated. The results show that in certain cases biases of unconfounded and confounded effects can result.