Effect size
An effect size states how large a difference or relationship is, in contrast to a p-value, which states only how unlikely the data would be if there were no effect at all. Common measures are Cohen's d, which expresses a difference between two groups in standard deviations, and the correlation coefficient r. The distinction matters because with a large enough sample almost any difference becomes statistically significant, so "significant" in a paper means "detectable", not "important". Reporting effect sizes with confidence intervals, rather than significance alone, is now standard practice in psychology journals.
Why it matters
Reading effect sizes rather than p-values is the fastest way to tell an important finding from a technically detectable one, in psychology and in reporting about it.
Also written: Cohen's d, effect sizes
Where it comes up
Read about Effect size on Wikipedia
Sources
- Effect size Wikipedia
- Statistical significance WikipediaThe thing effect size is usually contrasted with, and the source of most misreporting.
- p-value Wikipedia
