Evaluating Ordinal Survey Rating Scales: A Comparison of Methods

• In Event: 43.071 - Survey Research: Analyzing Items

Abstract

Researchers often administer surveys that contain ordinal rating scales assumed to reflect one or more constructs. An important step in analyzing data from these instruments involves determining the number of constructs present among the ordinal ratings. In this paper, we compared different methods for evaluating ordinal rating scales: (1) classical factor analysis, (2) ordinal factor analysis, and (3) Mokken scale analysis. Our results indicated that ordinal factor analysis rather than classical factor analysis or Mokken scale analysis resulted in a more-accurate determination of rating scale constructs. We recommend that researchers use appropriate statistical methods when analyzing ordinal rating scale data.

Authors

• Randall E. Schumacker, The University of Alabama

• Stefanie A. Wind, The University of Alabama - Tuscaloosa