Methods of Classification

This issue opens with a special section edited by Alexander von Eye, who sets out why classification deserves its own grouping of papers. The preface itself carries no abstract, but it frames what follows: a set of studies that each approach the problem of sorting people, behaviors, and measurements into meaningful categories from a different angle.

Raymond H. Baillargeon, Richard E. Tremblay, Doug Willms, Elisa Romano, Kyung Lee, and Hong-Xing Wu take on a problem that comes up constantly in developmental research. They ask how you model change within a single individual over time when there is no Gold Standard to measure that change against. Without a trusted yardstick, the modeling choices you make carry more weight than usual.

Edward W. Wolfe looks at the people doing the rating rather than the thing being rated. Using latent trait models, he shows how to pull out rater effects, the systematic tendencies that creep in when human judges score the same material differently.

Eduardo Gutiérrez-Peña contributes a treatment of Bayesian classification methods, bringing probability-based reasoning to the task of assigning cases to groups.

Von Eye returns, this time with Eun Young Mun and Alka Indurkhya, to consider developmental trajectories. Their framing is a decision-making one: typifying how a trajectory unfolds becomes a question of what you decide counts as one pattern versus another.

Patrick D. Converse and Frederick L. Oswald turn to applied territory, examining how the type of data you feed into a job classification changes both the classification itself and what it can be used for.

Christof Schuster digs into a measurement problem that is easy to overlook. When a dichotomous predictor contains measurement error, that error does not stay contained; it distorts the analysis of a continuous criterion in ways worth watching for.

Richard P. DeShon makes a pointed argument about group comparisons. Measures, he contends, cannot be treated as invariant across groups unless the error variances are homogeneous, a condition that often goes unchecked.

Von Eye closes the section with a piece on base models for Configural Frequency Analysis, laying out the foundational models the method rests on.