Interpretation of interaction in factorial analysis of variance design

Kenneth J. Ottenbacher

Research output: Contribution to journalArticlepeer-review

16 Scopus citations

Abstract

The validity of statistical conclusions in medical research depends on proper analysis and interpretation of collected data. One potential area of invalidity is the inappropriate post hoc analysis of statistically significant interactions in the analysis of variance of factorial designs. This paper examines the statistical explanations included in 83 studies published in three leading medical journals where the findings indicated significant interaction effects. Only 24 per cent of the reported statistically significant interactions had an accompanying correct interpretation. The most common form of misinterpretation involved a comparison of individual cell means within a row or column of one factor used in the design. This interpretation did not conform to the factorial ANOVA model with interaction. This misinterpretation occurs when the correct omnibus test of a hypothesis is followed by an incorrect post hoc analysis and/or an inaccurate assessment of the original statistical result.

Original languageEnglish (US)
Pages (from-to)1565-1571
Number of pages7
JournalStatistics in Medicine
Volume10
Issue number10
DOIs
StatePublished - Oct 1991
Externally publishedYes

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Fingerprint

Dive into the research topics of 'Interpretation of interaction in factorial analysis of variance design'. Together they form a unique fingerprint.

Cite this