Disattenuating Correlation Coefficients for Multiple Sources of Measurement Error, Scale Coarseness, and Congeneric Relationships

• In Event: Marriott Roundtable Session 35
  In Roundtable Session: Application of Factor Analysis

Abstract

We used confirmatory factor analysis (CFA) to evaluate effects of omission of key sources of measurement error, scale coarseness, and methods for modeling inter-item relationships (congeneric and essential tau-equivalent) on reliability and disattenuated correlation coefficients. One-thousand-seven-hundred-and-forty college students completed the Self-Description Questionnaire-III (SDQ-III, Marsh, 1992) on two occasions, a week apart. Results showed that psychometric properties of scores were most strongly affected by omission of a key source of measurement error (either transient or specific factor), followed respectively by scale coarseness, and misrepresentation of congeneric relationships. The most accurate reliability and disattenuated correlation coefficients were obtained from congeneric CFA models referenced to the continuous latent response variable metric.

Authors

• Walter Peter Vispoel, University of Iowa

• Carrie A. Morris, The University of Iowa

• Murat Kilinc, University of Iowa

• Mingqin Zhang, The University of Iowa