'RIGHT-SIZING' RESEARCH STUDIES: ASSURING ADEQUATE, NOT GROSSLY-OVERLARGE SAMPLE-SIZES

摘要

A "right-sizing" system is demonstrated that provides for: robust estimations of minimum study sample-size requirements, and efficient evaluations of alternate parameter assumptions. Importantly, 'assumption evaluations' often result in the avoidance of grossly-oversized studies [e.g., added $10K-$1000K costs]. "Right-sizing" calculations may be quickly conducted using simple-computational tools (eg, hand calculator) during a meeting as questions arise. Built upon an understanding of the robust mechanisms underlying the Dunlap "2-2" heuristic (re: α=0.05 2-tailed, and 1-β ≥ 0.80), it is applicable to the same range of dichotomous-to-interval data, but is also generalized to broadly accommodate correlated means and a full-spectrum of alternative α, 1-β combinations. We recommend our right-sizing system to HF/E and other practitioners interested in (1) Assuring adequate study sample-sizes, and (2) Avoiding grossly oversized studies that drain critical resources.