Visualizing Items and Measures: An Overview and Demonstration of the Kernel Smoothing Item Response Theory Technique

Abstract

Motivated by a renewed interest in exploratory data analysis and data visualization in psychology and social sciences, the current demonstration was conducted to familiarize a broader audience of applied researchers with the benefits of an exploratory psychometric technique – kernel smoothing item response theory (KSIRT). A data-driven, nonparametric KSIRT provides a visual representation of the characteristics of the items in a measure (scale or test) and offers convenient preliminary feedback about functioning of the items and the measure in a particular research context. The technique could be a useful addition to the analytical toolkit of applied researchers that work with a range of measures, within the classical test theory or IRT framework, and is suitable for use with a smaller number of items or respondents compared to parametric IRT models. KSIRT is described and its use is demonstrated with a set of items from a psychological well-being measure. A recently developed, easy to use R package was utilized to perform the analyses and the R code is included in the manuscript.