Reading focus
Syllabus: What Can Statistical Evidence Tell Us?
Compare contextual judgments of importance with conventional size labels. Keep the published correction about standardized units in view.
Why you’re reading it. State why an effect might matter before classifying it as small or large.
A question to consider. What would you need to know about a two-point gain on the school’s test to judge its importance?
Why this reading
Calling an effect small does not establish that it is unimportant. This paper supplies arguments for judging magnitude in context, while raising questions about when an effect’s consequences accumulate.
Before you read
A correlation describes a standardized association; its importance depends on the measures and consequences. The 2020 corrigendum corrects the original article: r equals the regression slope when both variables are standardized, not a slope in their original measurement units.
Read for
- Examine the premises behind using familiar effects as comparisons instead of treating a conventional size label as a verdict.
- Ask when repeated small effects might accumulate and when the context would prevent that accumulation.
- Distinguish effect magnitude, precision, reproducibility, and practical consequences.
Consider afterwards
- Two equally reliable associations have the same correlation: one concerns a minor inconvenience, the other a consequential outcome. What additional information would you need to compare their practical importance?
Editorial context
Use the published correction when interpreting the discussion of r and r². The paper challenges automatic judgments of importance; it does not make variance-based model comparisons illegitimate or show that every small association is useful. Its benchmarks and accumulation arguments are proposals to evaluate against the particular outcome, setting, and evidence.