On the Superior Statistical Properties of Frequency Scales in Job Analyses

Abstract

This study compared the statistical properties of four job analysis task survey response scale types: criticality, difficulty in learning, importance, and frequency. We used nine job analysis studies spanning two fields, medical imaging and allied health professionals, to compare the job analysis scales in terms of variability and interrater agreement. Results showed that frequency scales using absolute anchors had greater between-task variability and higher interrater agreement for all nine studies. This may have occurred due to what has been described by past research as self-presentation bias. In this case, an aggregate base percentage of respondents always responded that tasks in their domain are highly critical, highly important, and easy to learn. These results showed that frequency scales with absolute anchors yielded data with better statistical performance than other more subjective scales. These properties do not answer the question of whether a scale matches an exam's purpose, which is the most important consideration for job analyses. They do, however, suggest that, if statistics are a primary deciding factor, strong consideration should be given to using frequency scales with absolute anchors.