Effect Size in Education Research

Both are standardized mean differences — a group difference divided by a pooled standard deviation. The difference is that Cohen's d is biased upward in small samples, while Hedges' g applies a correction factor that removes most of this bias. Because educational studies often have modest samples, evidence standards such as the What Works Clearinghouse prefer the corrected g. The two converge as sample size grows, so for large studies the distinction is negligible.

They are useful rough conventions but are frequently misapplied. Cohen himself offered them tentatively, and in education many genuinely important, well-implemented interventions produce effect sizes well below 0.5, especially on broad standardized tests. What counts as a meaningful effect depends on the outcome, the comparison condition, the cost, the duration, and the population. Interpreting effect sizes mechanically against generic thresholds, rather than against context and comparable studies, is a common error.

Because a p-value confounds the size of an effect with the size of the study. A trivial effect can be highly significant with a large sample, and an important effect can be non-significant with a small one. The effect size separates magnitude from sample size, telling you how much something changed, and its confidence interval communicates the precision of that estimate. Reporting standards and meta-analysis both require effect sizes precisely because significance alone cannot support decisions about practical importance.