Regression discontinuity designs in a latent variable framework.

Abstract

When randomized control trials are not available, regression discontinuity (RD) designs are a viable quasi-experimental method shown to be capable of producing causal estimates of how a program or intervention affects an outcome. While the RD design and many related methodological innovations came from the field of psychology, RDs are underutilized among psychologists even though many interventions are assigned on the basis of scores from common psychological measures, a situation tailor-made for RDs. In this tutorial, we present a straightforward way to implement an RD model as a structural equation model (SEM). By using SEM, we both situate RDs within a method commonly used in psychology, as well as show how RDs can be implemented in a way that allows one to account for measurement error and avoid measurement model misspecification, both of which often affect psychological measures. We begin with brief Monte Carlo simulation studies to examine the potential benefits of using a latent variable RD model, then transition to an applied example, replete with code and results. The aim of the study is to introduce RD to a broader audience in psychology, as well as show researchers already familiar with RD how employing an SEM framework can be beneficial.

Guidelines to improve internationalization in the psychological sciences

“…Van de Vijver (2013) discusses how there are strong expectations of editors and reviewers regarding what should be included in manuscripts and where or what statistical techniques are appropriate; if the information presented does not conform to these implicit norms, the authors may be misconstrued as incompetent. For instance, there is an implicit norm of conducting laboratory experiments (but not necessarily field studies) in social psychology and accepting that lab experiments are the best way to determine causation (Grosz et al., 2020), whereas in another field, regression discontinuity designs (e.g., Soland et al., 2022) might be acceptable. Similarly, though statistical mediation analyses (Rucker et al., 2011) were popular in experimental social psychology a few years ago, they fell out of favor in recent years.…”

Limitations of the Sum-and-Alpha Approach to Measurement in Behavioral Research

“…If the data are poor-as when there is no validity evidence supporting scores-the most sophisticated data analysis will remain susceptible to incorrect conclusions and inaccurate estimates (Flake, 2021). Weaknesses of unsubstantiated sum scoring have been demonstrated with advanced statistical approaches such as regression discontinuity (Soland et al, 2022), machine learning (Jacobucci & Grimm, 2020), intensive longitudinal data and time-series analysis (McNeish et al, 2021), growth modeling (Kuhfeld & Soland, 2022), network modeling (Haslbeck et al, 2022), and clinical trials (Kessels et al, 2021). More simply, the quality of the statistical model is limited by the quality of the data itself and no statistical workaround can avoid consequences of poor measurement.…”