RESEARCH DESIGN ISSUES IN GROUPING-BASED TESTS
成果类型：
Article
署名作者：
LYS, T; SABINO, JS
刊物名称：
JOURNAL OF FINANCIAL ECONOMICS
ISSN/ISSBN：
0304-405X
DOI：
10.1016/0304-405X(92)90033-T
发表日期：
1992
页码：
355-387
关键词：
摘要：
With grouping, a sample is sorted by an observable variable and the mean values of the dependent variable in the extreme-ranked groups are compared. We show that test power is maximized when the two extreme groups each contain 27% of the sample, a much larger percentage than that typically used in the literature. This result is not sensitive to the distribution of the dependent variable. We also show that regression is unambiguously more powerful than grouping, even when the independent variable is measured with error.
来源URL：