Non‐Standard Log‐Linear Models for Orthogonal Prediction Configural Frequency Analysis

The paper discusses L AUTSCH , L IENERT . and VON E YE'S (1992) orthogonal prediction configural frequency analysis (PCFA). Problems with this approach result from a lack of rules for specification of orthogonal comparisons, and because it is not obvious which comparisons are orthogonal. As an alternative it is proposed estimating non‐standard log‐linear models. These models specify orthogonal PCFA comparisons via orthogonal effect coding variables. This approach kadvantageous in that it allows researchers to simultaneously test orthogonal comparisons and PCFA parameter estimates.