Solved – use a partial eta squared on t-test results

Question

I am currently trying to decide which effect size is the most appropriate for my results. I have conducted a repeated measures ANOVA, there are five variables in the ANOVA, all of which are within samples.

I have used the partial eta squared value as the effect size for the overall model. However, my results also rely quite heavily on the pairwise comparisons between the five variables so I need to report the effect sizes for these. -But ANOVAs don't give an effect size for these within-subject comparisons. They do actually report the R but this is not appropriate for this.

I need the effect sizes for these comparisons and I am not sure what to use- I have conducted paired samples t-tests and worked out the partial eta squared. I have also conducted multiple RM ANOVAs for each pairing to check that the partial eta squared was correct from the t-test (it was).

So I guess my question is- is it acceptable to report the a t-test and partial eta squared or does this have to Cohen's d instead?
Or should I state that i conducted lots of RM ANOVAs to get these effect sizes?

I am reluctant to use Cohen's d as I wanted to be consistent in my use of effect size. I also understand that Cohen's d sometimes has problems with conducting the effect size of paired samples. Lastly, this research is for my PhD- which is in education- and I am not sure whether Cohen's d would be as meaningful to my examiners. I also prefer partial eta squared as it is on a scale from 0-1.

Answer 1

Want to improve this post? Provide detailed answers to this question, including citations and an explanation of why your answer is correct. Answers without enough detail may be edited or deleted.

I believe that you can apply the formula

pEta^2 = t^2/(t^2 + df)