Sometimes t-values with p-values are reported as effect estimates. Can they be used for meta-analysis based on standardized mean difference?
Hello Martin,
While t-values may be used by some, they are not very good effect size estimators, since they depend on sample size. Effect size estimates should, ideally, be independent of sample size.
There are simple conversions for test statistics such as t- or F- to common effect sizes available. Here are some sources:
https://www.psychometrica.de/effect_size.html
https://www.polyu.edu.hk/mm/effectsizefaqs/calculator/calculator.html
Good luck with your work!
You can convert a t-value to a standardized mean difference if you now the sizes of the groups using the following equation.
ES = t * square root ((n1 + n2)/n1n2)
Hallo Martin,
in addition to what the others suggested, you can have a look to the formulas in Appendix C of the following document - it provides a useful summary of formulas for effect sizes conversion: https://campbellcollaboration.org/media/k2/attachments/converting_between_effect_sizes.pdf
Just updating Gabriele Midolo's great answer, since the link he provided, is currently unavailable.
Here is a web-based effect-size calculator:
https://www.campbellcollaboration.org/research-resources/effect-size-calculator.html
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis