STATA has recently implemented effect size calculations for regression models in their postestimation procedures measures of effect size in Stata 13. It takes a regression or ANOVA object as input, and returns eta-squared or omega squared as well as their confidence intervals.

Is anyone aware of any such or similar implentation in R software, or how this could be achieved using the contents of lm or aov objects?

  • $\begingroup$ Did you find a solution for this problem? I have the same problem: I want to calculate CIs for eta sq. in an ANOVA with R. $\endgroup$ – Deleet Nov 1 '15 at 21:56

I believe that your best bet is to use MBESS R package. For more details, please see the package's home page, this comprehensive JSS paper and the package's home page on CRAN.

In addition, consider package bootES for calculating bootstrap effect sizes: package's CRAN page. This relevant tutorial might also be helpful. Finally, I highly recommend this excellent article, which not only provides general guidelines on calculating and interpreting effect sizes and confidence intervals, but also some specific recommendations for studies with different research designs.

  • $\begingroup$ Aleksandr, thanks for your suggestion. How would you go about obtaining eta-squared with confidence intervals using the package, papers and tutorials you mentioned? $\endgroup$ – Tormod Jan 20 '15 at 7:09
  • $\begingroup$ @Tormod: You're welcome. I will try to come up with an example. $\endgroup$ – Aleksandr Blekh Jan 21 '15 at 2:34
  • $\begingroup$ Could you come up with an example? $\endgroup$ – Deleet Nov 1 '15 at 23:29
  • $\begingroup$ @Deleet: Sorry, I don't have time for this now (and not sure whether I have enough expertise as well). I suggest you to read the article and tutorial, linked above. $\endgroup$ – Aleksandr Blekh Nov 2 '15 at 17:55
  • $\begingroup$ I have gone thru tons of these and various R packages too. The MBESS package produces wrong results I think since the CIs do not contain the point estimate when CI is <50%. If not, then it is using an odd approximation that doesn't work for low CIs. $\endgroup$ – Deleet Nov 2 '15 at 21:21

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.