Interaction Fallacy
Definition: A statistical error in which a formal test is not conducted to assess the difference between a significant and non-significant correlation (or other measures, such as Odds Ratio). This fallacy occurs when a significant and non-significant correlation coefficient are assumed to represent a statistically significant difference but the comparison itself is not explicitly tested.
Related terms: Comparison of Correlations, Null Hypothesis Significance Testing (NHST), Statistical Validity, Type I error, Type II error
References:
- Gelman, A., & Stern, H. (2006). The difference between âsignificantâ and ânot significantâ is not itself statistically significant. The American Statistician, 60(4), 328â331. https://doi.org/10.1198/000313006X152649
- Morabia, A., Have, T. T., & Landis, J. R. (1997). Interaction Fallacy. Journal of Clinical Epidemiology, 50(7), 809â812. https://doi.org/10.1016/S0895-4356(97)00053-X
- Nieuwenhuis, S., Forstmann, B. U., & Wagenmakers, E. J. (2011). Erroneous analyses of interactions in neuroscience: a problem of significance. Nature Neuroscience, 14(9), 1105â1107. https://doi.org/10.1038/nn.2886
Originally drafted by: Graham Reid
Reviewed by: Ali H. Al-Hoorie, Mahmoud Elsherif, Kai Krautter, Sam Parsons