Development of a Revised Deviance Analysis Approach


August 2022


Ki Ho Kim, Matthew C. Reeder, Jessica L. Harris, Jeffrey A. Dahlke

Phil Lewis
National Center for O*NET Development


The O*NET Program conducts data cleaning and preparation activities as part of its annual Job Incumbent and Occupational Expert (OE) data collection update efforts. An important part of this process is identifying respondents that are not representative of their sampled and targeted occupation (i.e., deviant respondents). A two-stage deviance analysis identifies and removes respondents whose task ratings and other information suggest they are not representative of their assigned occupation.

This paper describes the work to update and streamline the deviance analysis process. The new process will be implemented by Human Resources Research Organization (HumRRO) staff each year as a standard cyclical activity.