Table-Specific Continuity Corrections for Configural Frequency Analysis

This paper introduces a set of continuity corrections built for six of the tests used in Configural Frequency Analysis (CFA). The idea is simple at the level of a single table: each table gets one correction, and that correction takes the form of a constant.

What determines the size of that constant is not fixed in advance. It shifts according to several features of the table itself, among them the degrees of freedom, the overall size, the cell frequency, the strength of type, and the nominal level a. Strength of type here is treated as a quantitative way of describing types rather than a qualitative label.

The authors tested these corrections through simulation. Out of the six tests, three accounted for more than 90% of the best solutions: the test put forward by Perli and colleagues, the test newly proposed by Lautsch and von Weber, and the c2-component test.

The work is credited to Stefan von Weber, Erwin Lautsch, and Alexander von Eye.