Simpson's paradox: An example using accident data from the state of Texas

✍ Scribed by Keith Smith; James O'Day

• Publisher: Elsevier Science
• Year: 1982
• Tongue: English
• Weight: 152 KB
• Volume: 14
• Category: Article
• ISSN: 0001-4575
• DOI: 10.1016/0001-4575(82)90079-3

✦ Synopsis

Al~traet--Two-way tabulation of data when a third interacting variable is present can lead to false inferences. In this paper it is shown that the probability of an occupant fatality given a crash in a single-vehicle accident is independent of vehicle size in an aggregate data set. When the data set is partitioned into 2 occupant age groups, it is seen that both of the age groups have higher fatality rates in smaller cars. This is discussed as an example of a paradox presented by Simpson in his 1951 paper.