Political scientists often claim variables have no effect when coefficients aren't statistically significant. This article introduces researchers to more robust methods, such as 90% confidence intervals, for demonstrating negligible impacts and supporting their hypotheses with stronger evidence.
Key Concept: Statistical Significance vs. Practical Negligibility
The common approach of relying solely on p-values (especially the default cutoff) is insufficient for proving true negligibility.
New Approach: Using 90% Confidence Intervals
Instead, researchers can employ narrower confidence intervals like the 90%, providing more compelling evidence that effects are genuinely small or zero.
Supporting Examples: Illustrative Cases
Several examples demonstrate how shifting from reliance on weak statistical tests to these stronger interval-based methods clarifies findings and strengthens arguments for negligible political influence.
| Arguing for a Negligible Effect was authored by Carlisle Rainey. It was published by Wiley in AJPS in 2014. |
