Calculate Hedges g Without Losing the Outcome Meaning

Decide what should be standardized

Start with the construct, not the formula. Record what each instrument measures, its scoring range, and whether a high score means better or worse status. Standardizing two unrelated outcomes does not make them interchangeable. Write an inclusion decision for each instrument before inspecting the resulting effect sizes.

For a two-group comparison, Hedges g expresses the difference in means in pooled standard-deviation units, with a small-sample bias correction. The metafor effect-size documentation describes standardized mean differences, required inputs, and bias correction.

Assemble a complete extraction row

Record the mean, standard deviation, and analyzed sample size for each group at the chosen time point. Check that a reported uncertainty measure is actually an SD, rather than an SE or confidence interval. Keep the original outcome label, units, page reference, and any transformation in the continuous-outcome worksheet.

Do not mix baseline sample sizes with follow-up means when attrition changes the analysis population. Keep endpoint scores and change scores explicitly labeled. A paired change calculation needs information that an independent-groups calculator does not recover from the endpoint summaries alone.

Recompute a fictional example

Suppose two independent groups each have 40 participants. Group one has mean 10 and SD 4; group two has mean 12 and SD 4. The pooled SD is 4, and the group-one-minus-group-two standardized difference is −0.50.

With 78 degrees of freedom, the approximation J = 1 − 3/(4 × 78 − 1) gives J about 0.990354. Multiplying gives Hedges g approximately −0.49518. The Hedges g calculator uses this approximate correction. Software using an exact correction can differ slightly. These invented data illustrate arithmetic only.

Interpret direction before magnitude

If higher scores mean worse symptoms, the negative value favors group one on that outcome. If higher means better function, the same sign has the opposite interpretation. Resolve this with explicit outcome labels; do not rely on color or a minus sign alone.

Avoid translating a standardized difference into a universal clinical benefit category. The scale, population, observed variability, and clinical context still matter. Record whether an adjusted estimate would better match the review question, and do not substitute this crude independent-groups calculation for clustered, paired, or repeated-measures analysis.

Your final extraction record should let a colleague reproduce both the numeric result and the decision about which direction favors which group.

Source checked on 30 October 2026; numerical example independently recalculated.