Difficulties in the statistical control of halo

Abstract

Criticizes an article by F. Landy et al (see record 1981-00274-001 ) that suggested a method for statistically controlling halo error in ratings that involved partialing out an overall evaluation from specific performance appraisal items. It is argued that this implies that specific items ought to be unrelated to overall performance judgments and that observed correlations are the result of rater errors. It is also suggested that the Landy et al method inevitably removes some true score variance in ratings. The procedure is also criticized on the basis of an inappropriate unit of analysis. (24 ref) (PsycInfo Database Record (c) 2025 APA, all rights reserved)