Impact of Design Features for Cross-Classified Random-Effects Logistic Models When the Cross-Classification Structure Is Ignored: A Monte Carlo Study

• In Event: Poster Session 19
  In Poster Session: Multilevel Modeling Methods and Applications IV

Abstract

The consequences of ignoring the existence of cross-classified data structures for continuous outcomes have been investigated in recent research. However, little work has been conducted to examine these consequences for similarly structured datasets with dichotomous outcomes. This study is based on analysis of recidivism of incarcerated youth, cross-classified within counties of origin and treatment facility, and uses Monte Carlo simulation to investigate factors affecting the quality of cross-classified models for dichotomous outcomes. Our study focuses on how the correlation of level-two factors’ residuals, number of facility feeders, number of levels of cross-classified factors, intra-unit correlation coefficient (IUCC), and parameter estimation methods affect the fixed and random effects and model fit when the cross-classified data structure is appropriately modeled versus ignored.

Authors

• Weijia Ren, The Ohio State University

• Ann A. O'Connell, The Ohio State University

• William Loadman, The Ohio State University

• Raeal Moore, ACT, Inc