2
$\begingroup$

I want to calculate effect sizes for running a meta-analysis. I am interessted in the difference between two groups after a treatement. In the papers included for the meta-analysis I often find the mean and standard deviation for each group, as well as (for example) t-values. In this case, the estimated effect size for the study varies - depending if I use the means and standard deviations for calculation of the t-test.

Which one would be the appropriate estimate - the one calculated based on means or the one calculated based on the t-value?

KR

$\endgroup$
6
  • $\begingroup$ Can you edit to give us some more detail? At the moment it is not clear what calculations you are doing to derive the standardised mean difference via either route. $\endgroup$ – mdewey Nov 24 '20 at 12:12
  • $\begingroup$ Sure (and thanks for responding). $\endgroup$ – Martha Nov 24 '20 at 12:55
  • $\begingroup$ I would like to calculate Hedge g from the mean, standard deviation and group size. I use the package esc in R to calculate effect sizes. $\endgroup$ – Martha Nov 24 '20 at 12:56
  • $\begingroup$ Its just that I have sometimes the mean, standard deviation and group size and additionally the authors also report the t-values. It is possible to calculate Hedge g via mean, standard deviation and group size as well as thorugh the reported t-value. So I was wondering which values to choose for effect size calculation. $\endgroup$ – Martha Nov 24 '20 at 12:57
  • $\begingroup$ They should give the same value although if they are not reported with many decimal places they may differ due to rounding error. $\endgroup$ – mdewey Nov 24 '20 at 13:26
1
$\begingroup$

Sample size, mean, and standard deviation should be prioritized over other statistics (e.g., t-test or ANOVA values) to estimate standardized mean differences (e.g., Hedges' g).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.