0
$\begingroup$

I have two groups (experimental and control) and a measure (an estimate of confidence) measured at baseline and at post-intervention. I have ran a multiple regression and found that the intervention condition (experimental or control) has a significant effect on baseline to post-intervention confidence. Based on some other StackExchange posts, I calculated Cohen's f-squared as a measure of effect size. However, I also need to calculate the Cohen's d because it is the more commonly reported measure of effect size. To calculate Cohen's d, do I compare baseline of experimental vs. post_intervention of experimental OR (baseline-post_intervention) of experimental vs. (baseline-post_intervention) of control? I realize the question indicates a lack of basic understanding, but I could really use some help.

$\endgroup$
4
  • 1
    $\begingroup$ You could calculate Cohen's d for whichever pairs of groups you want. However, there's some discussion about how to calculate Cohen's d for groups from a larger regression model, as the standard error from the analysis isn't simply the standard deviation from any of the groups. $\endgroup$ Feb 6, 2023 at 17:52
  • $\begingroup$ Can you please share the literature? $\endgroup$
    – Outsider
    Feb 6, 2023 at 18:06
  • 1
    $\begingroup$ I don't have any specific recommendations for publications on this, but you might start with this discussion: stats.stackexchange.com/questions/603792/whats-the-appropriate-effect-size-estimate-and-power-analysis-for-post-hoc-re. ... As mentioned there, if you are using R, and a general linear model, the eff_size() function in the emmeans package calculates a Cohen's d -like effect sizes from the model. $\endgroup$ Feb 6, 2023 at 19:25
  • $\begingroup$ Thanks a lot !! $\endgroup$
    – Outsider
    Feb 6, 2023 at 19:40

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.