Latent Variable Modeling and Psychometric Methods in HRM

Measuring and analyzing latent (i.e., non-observable) constructs or traits is a fundamental issue in quantitative research. During the last 50 years, numerous statistical models have been developed for this purpose. Initially, these models were used more or less exclusively in psychological diagnostics. Due to their outstanding statistical and substantial properties, a strong interest outside the application area of social sciences has emerged recently. Such fields of research are, for instance, medicine, marketing, and human resource management (HRM). Particularly, the HRM field provides highly relevant application areas such as large-area screening for admission restrictions to, e.g., universities. Consequently, new methodological and computational challenges arise. From the methodological point of view the focus is on enhancements on already existing models in Item Response Theory (IRT), Multidimensional Scaling (MDS), and Structural Equation Models (SEM). From a statistical point of view it is desirable to embed these models into larger frameworks of model families. Hence, common mathematical routines for parameter estimation can be applied offering numerous aspects of model flexibility. Another methodological aspect of this project is the elaboration of connections between already existing models, e.g., between IRT and SEM. For the practical use these methodological improvements lead to a more detailed and more flexible way to measure, analyze, and interpret models with latent variables involved. These new improvements as well as already existing models must be computationally accessible to researchers of a wide field of substantive areas. Additionally, the corresponding implementations have to be flexible enough to allow user customizations such as further innovative analyses, result representations, visualizations, etc. Hence, a main part of this project is developing packages for the well-known open-source environment of statistical computing "R" which provides the claimed usability and flexibility. The corresponding routines must be capable processing large amounts of data since in HRM, like in other fields, data can become considerably large. At UCLA, there are two leading experts in the corresponding field of psychometrics and applied statistics with latent variables, as well as computational statistics: Jan de Leeuw and Peter Bentler. Prof. de Leeuw is author of numerous articles in the field of IRT and MDS, has been contributing to many R packages, and founded the PsychoR project for psychometrics in R. Prof. Bentler is one of the leading pioneers in SEM and latent variable modeling. Their huge experience and corresponding competence is indispensable for methodological developments and efficient computational implementations to enable the application of these models in any field where latent variables are an issue.