Theory-Based Directional-Testng: Revisiting a Primary Option for Efficient Human-Factors/Ergonomics (HF/E) Studies
Alvah C. Bittner, Frank J. Winn, Rachel C. L. Bittner, N. Catherine Lundy · Proceedings of the Human Factors and Ergonomics Society Annual Meeting · 2004
This study revisited use of directional-models/analyses in HF/E. Surveys revealed both that: these models/analyses are not widely used in studies reported in HF/E journals, and practitioners avoid them due to misconceptions regarding their scientific validity. For applications-orientated HF/E studies, two conditions are demonstrated where directional testing is appropriate: (a) When unpredicted direction differences would be meaningless (e.g., cumulative-risk declines with increased exposure); and (b) When unpredicted directional findings will result in the same action as no difference (e.g., many programand product-evaluations). One-vs. two-tailed testing results in sample-size savings of æ30% across virtually all situations where the Student t-test or corresponding ‘ directional F-test’ may be appropriately applied. HF/E practitioners are encouraged to: Utilize onetailed and directional models/analyses when conditions reflect the logical or functional ‘one-sidedness’ implicit in many HF/E studies, and Take full advantage of the substantially reduced sample-sizes (æ70%) and associated resource expenditures.