Scale-Point Effects on Reliability and Disattenuated Correlation Coefficients Derived From Confirmatory Factor Analysis Raw-Score and Continuous-Latent-Response-Variable Models

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

Abstract

We used confirmatory factor models (CFAs) referenced to both raw-score (RAW) and continuous-latent-response-variable (CLRV) metrics to investigate effects of number of scale points on indices of reliability and validity after controlling for multiple sources of measurement error. Data were collected from over 1,700 respondents who completing a multidimensional self-concept on two occasions. On average, the CLRV models provided higher reliability and disattenuated validity coefficients than did the RAW models. Differences between the CLRV and RAW models were more pronounced when moving from two to four response options than when moving from four to eight response options. These findings underscore the usefulness of the CLRV-CFA model across situations and particularly when using items that are dichotomously scored.

Authors

• Carrie A. Morris, The University of Iowa

• Walter Peter Vispoel, University of Iowa

• Murat Kilinc, University of Iowa