When Interactions Bias Selected Corrections

• In Event: Marriott Roundtable Session 29
  In Roundtable Session: Missing Data Across Statistical Models

Abstract

: Corrections for range restriction (e.g., Case II and Case III) assume a linear relationship between the predictor of interest and the criterion. This will generally not be the case when interaction effects exist (e.g., between the direct and the incidentally selected predictor variables). In this paper, I investigate through Monte Carlo simulation the degree to which interaction effects bias estimates corrected using Case III. I also develop a correction that generally mitigates bias stemming from interactions and conclude with recommendations on how researchers can identify interactions and correct for range restriction.

Author

• Dustin Fife, Rowan University