作者: Ramesh Chandra , Kermit Rohrbach , G. Lee Willinger
DOI: 10.1111/J.1540-6288.1995.TB00852.X
关键词:
摘要: This paper distinguishes between testing approach (parametric t, sign, cross-sectional rank, and longitudinal rank) test type (conventional, standardized, weighted least squares). Similar tests (i.e., of the same type) are compared for each using constant unequal shift models. Thus results given here free confounding effects present in prior comparisons dissimilar tests. While all sensitive to underlying location model, if is independent security standard deviation, then squares rank most. powerful.