Responding to Low Coefficient Alpha: Potential Alternatives to the File Drawer

Abstract

To the detriment of human resource development (HRD) theory building and research, many scholars may think that research data with a low coefficient alpha is destined for the file drawer; this does not have to be the case. Contemporary literature suggests that many scholars do not know how to move forward with data that yields α < .70. In addition, an investigation revealed that many scholars practice the method of item deletion to increase alpha. Besides supporting the case that discarding research simply because of low coefficient alphas may be unnecessary, a guide is presented to demonstrate how scholars and scholar–practitioners may be able to analyze data when an initial estimate of internal reliability is low. We caution that deleting items may increase reliability at the cost of validity. As an alternative, this study demonstrates that eliminating subjects can increase alpha and maintain the integrity of the scale. This guide presents generalizability theory as a means to identify the source of error variance in data as well as a step-by-step process to correct for low coefficient alpha. The guide is illustrated with data and R syntax.