You have the <em>t</em> and the journal wants an effect size. The conversion is exact, but it needs <strong>both group sizes</strong>: the shortcut formula going around online only holds when the groups are equal.
d = t · √(1/n₁ + 1/n₂)
d is the mean difference over the pooled standard deviation. t is that same difference over its standard error, and the standard error carries each group size inside it. When you rearrange, the sizes do not cancel: they remain as √(1/n₁ + 1/n₂).
When both groups are the same size that factor simplifies to 2/√N, which is where the short formula comes from. With unequal groups it stops simplifying, and using it anyway inflates the effect.
The numbers: with t = 2.5 and a hundred people split 50-50, d is 0.50 either way. Split 90-10, the exact formula gives 0.83 while the shortcut still says 0.51. That is not a decimal, that is a different conclusion.
d is upward biased in small samples. The Hedges correction fixes it by multiplying by J = 1 − 3/(4·df − 1), which is why g always comes out slightly smaller.
At 200 degrees of freedom J is 0.996 and the difference is cosmetic. At 8 it is 0.904: d was overestimating by 10 %. Below twenty per group, reporting d when you could report g is a reviewer comment waiting to happen.
Many psychology and biomedical journals now ask for g by default, and every serious meta-analysis works with it.
Its confidence interval. A d = 0.50 with interval [0.05, 0.95] and one with [0.42, 0.58] say very different things, and the first is what small samples usually give you.
And the direction. The sign of d depends on which group you subtracted first; saying which avoids half the misunderstandings in a discussion.
How to write all of it in APA 7 is in how to report Cohen’s d, and if what you are missing is the interval, the interval against its estimate shows how to recover the standard error.
All thirteen effect size conversions, grouped by what each one is for, are in the index.
If you use it in your thesis or in a paper, cite it so that readers of your Methods section know how the calculation was done.
Ciudad-Fernández, V. (2026). t to Cohen's d converter [Web application]. Ciudad Science. Retrieved October 14, 2026, from https://analisisdedatospsicologia.com/en/convert/t-to-cohens-d
For your Methods section: Cohen's d was obtained from the t statistic and the group sizes, using the Ciudad Science t to Cohen's d converter (Ciudad-Fernández, 2026).
@misc{ciudadfernandez2026convtaddecohen,
author = {Ciudad-Fern{\'a}ndez, V{\'i}ctor},
title = {{t to Cohen's d converter}},
year = {2026},
publisher = {Ciudad Science},
howpublished = {Web application},
url = {https://analisisdedatospsicologia.com/en/convert/t-to-cohens-d},
urldate = {2026-10-14},
note = {Accessed 2026-10-14},
}Not directly. In repeated measures there are two possible ds — one using the standard deviation of the differences, one using the original scores — and they give very different numbers. This page handles the independent-samples case.
Only for direction. Convert the absolute value and then attach the sign that matches the comparison you are describing. What matters is stating which group is which.
The conversion is no longer exact, because Welch does not use a pooled standard deviation. You can give it as an approximation and say so, but the clean route is computing d from each group’s mean and standard deviation.
Cohen’s (.20, .50, .80) are general references he himself offered as provisional. In a field with naturally small effects, .30 can be large. What gives the number meaning is comparison with your own literature, not the table.