Multiple Scale, Multiple Categories: Investigating Potential Bias With Robust Estimation

• In Event: 61.084 - Criteria and Bias in Measuring Latent Constructs and Classes

Abstract

Recently, advances related to estimation techniques to analyze coarsely categorized, non-normal data have been mad (e.g., robust ML and Diagonally Weighted Least Squares methods, Bayesian estimation) Such techniques have showed promise for the analysis of ordinal data and have led researchers to suggest that strategies such as parceling variables to achieve more “continuous-like” data are not needed. One area which has not received attention is how data from multiple scales, where scales have different numbers of scale points, may affect estimation. This study will use simulation to compare the effectiveness of different robust estimation methods, when data arise from scales with different numbers of categories but are analyzed within the same model.

Authors

• Grant B. Morgan, Baylor University

• Christine DiStefano, University of South Carolina

• Elizabeth Leighton, University of South Carolina