OPTIVAL: An R package for model-based optimal test validity analysis

Identificador:  imarina:9681566

Handle: https://hdl.handle.net/20.500.11797/imarina9681566

Autores:  Navarro-González D; Ferrando PJ; Morales-Vives F

Resumen:
Maximizing external validity (the capacity of test scores to predict relevant criteria) is a central concern in psychological and educational measurement. This article presents OPTIVAL, an R package that implements a model-based approach for selecting item subsets that optimize the predictive validity of test scores with respect to an external criterion. Items are calibrated with a unidimensional factor-analytic model, model-implied validity curves and optimal item rankings are derived, and the accuracy of the predictions is checked through bootstrap resampling. The package provides both programmatic R functions and an interactive Shiny application, and is freely available on GitHub under a GPL-3 license. Its functioning is illustrated with simulated and real data examples.