Methods matter: Testing competing models for designing short-scale Big-Five assessments

Many psychological instruments are psychometrically inadequate because derived person-parameters are unfounded and models will be rejected using established psychometric criteria. One strategy towards improving the psychometric properties is to shorten instruments. We present and compare the following procedures for the abbreviation of self-report assessments on the Trait Self-Description Inventory in a sample of 14,347 participants: (a) Maximizing reliability/main loadings, (b) Minimizing modification indices/cross loadings, (c) the PURIFY Algorithm in Tetrad, (d) Ant Colony Optimization, and (e) a genetic algorithm. Ant Colony Optimization was superior to all other methods in improving the model fit of short scales. We conclude that in lengthy inventories Ant Colony Optimization currently represents the most purposeful and versatile method for developing psychometrically sound brief personality scales.