Hi, I am currently in the process of writing up my results from a research project. The assumptions for running a Manova were violated so I used a series of Kruskal Wallis tests using SPSS.
I have then looked at pairwise comparisons as a follow-up, however, I am unsure of how to report this, as the output provides a test statistic and a significance value. Would I need to calculate effect sizes for this? If so, which method would be best? The Andy Field SPSS book advised calculating an effect size r - is this referring to Pearson's r, as this is not very clear?
Thank you in advance!
What you did is not entirely clear but it sounds like you have a lot of comparisons. Given that, look at this reference(attached). I would consider using one or another of these suggestions depending on your exact situation. Best, David Booth
Kindly check the book "Discovering statistics using IBM SPSS" by Andy Field.
Hello Aamna,
Marascuilo and Serlin (1988), Statistical methods for the social and behavioral sciences, give this relationship between eta-squared (variance accounted for) and the Kruskal-Wallis H statistic (a chi-square):
eta-squared = H / (N - 1)
Which looks like the usual conversion of a chi-square test of independence to a phi coefficient (except you'd take the square root of H / (N -1) ).
The eta-squared may be used as an R-squared substitute; it's square root may be used as a phi or correlation coefficient substitute.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS