Cross-classified models in I/O psychology

Cross-classified models accommodate data structures that have more than one cluster variable, which are not nested in each other but overlap. They simultaneously consider all clustering variables. This allows one to study effects on several levels at once. Cross-classified data structures are common in various field of applied research (e.g., research on teams, career paths, interventions). The present article demonstrates modeling options and specifications for cross-classified data to be used within these different research strands. For more specific demonstration purposes, we use a data set on rater variance in assessment centers. Commonly, raters observe only a subset of participants and hence ratings are both nested in participants and raters, but participants and raters are not nested in each other. Using cross-classified models allows studying sources of rater variance (e.g., professional expertise) and interactions between cluster level variables, for example, interactions between participants' and raters' personality and sociodemographic characteristics (e.g., gender). From a practical research point of view and to ease application, we deliver a step-by-step overview of modeling procedures and power analysis including software code. •Cross-classified models as a powerful approach to hierarchical data with overlapping clusters•Numerous examples for applications of cross-classified models in I/O research•Empirical demonstration with cross-classified data from an assessment center•Step-by-step overview of modeling procedures and power analysis including software code