期刊名称：Practical Assessment, Research and Evaluation
印刷版ISSN：1531-7714
电子版ISSN：1531-7714
出版年度：2010
卷号：15
出版社：ERIC: Clearinghouse On Assessment and Evaluation
摘要：Yule (1903) and Simpson (1951) described a statistical paradox that occurs when data is aggregated. In
such situations, aggregated data may reveal a trend that directly contrasts those of sub-groups trends.
In fact, the aggregate data trends may even be opposite in direction of sub-group trends. To reveal
Yule-Simpson’s paradox (YSP)-type occurrences, researchers must simultaneously consider the effect
of an intervention at specific levels and on the overall model to ensure datasets are accurately analyzed
and research findings are appropriately interpreted. The primary objectives of this manuscript are to:
(1) examine the history of YSP; (2) describe necessary and sufficient causes for YSP occurrences; (3)
provide examples of YSP in research and explain YSP’s relationship to multi-level modeling including
Hierarchical Linear Modeling (HLM); and (4) discuss YSP’s implications for researchers
