Control of Type 1 and Type 2 Errors in Configural Frequency Analysis

Abstract

In this article, standard configural frequency analysis (CFA) is compared with combinatorial Victor CFA with respect to Type I errors and power. Victor CFA is most important and suitable when phantom types or phantom antitypes are suspected to exist. These can emerge when types and antitypes contain cases from populations other than the sample under study, and, thus, distort marginal probabilities. Simulation results are presented that suggest that, when samples are comparatively small, standard CFA has more power. In contrast, Victor CFA has more power when sample sizes increase. The concept of type strength is introduced. It can be used as a supplement to statistical CFA testing. Data examples illustrate that the two variants of CFA can result in dramatically different data evaluations. In the article, it is discussed that Victor CFA can also be used when base models are more complex than the main effect model that was used in Lienert’s standard CFA and Victor’s original alternative CFA.