0
$\begingroup$

I'm running a two way repeated measures ANOVA, and I'm having trouble finding what is the appropriate output to report (Contrasts vs Pairwise Comparisons).

I have a 2x5 design. Lets say I have Variable 1(group 1 and 2) and Variable 2 (groups a,b,c,d,e). I want to assess whether there is a difference in means based on separating data by variable 1 (group 1 vs group 2), variable 2 (a vs b vs c vs d vs e), or an interaction of both (1a vs 1b..etc).

However, I am haven't found much written on whether within-subject contrasts or pairwise comparisons are more appropriate to report after I found significant differences.

For instance, I found that means are significantly different between groups a,b,c,d,e. However, I now want to know WHERE the difference is (for instance is a larger b, etc).

Thank you for the help.

$\endgroup$
  • $\begingroup$ Then you are looking for post-hoc comparisons, which is well covered on this site. $\endgroup$ – chl Oct 31 '20 at 19:25
0
$\begingroup$

Functionally, they operate similarly.

From this page (demo in SPSS)

http://www.uvm.edu/%7Edhowell/StatPages/More_Stuff/RepMeasMultComp/RepMeasMultComp.html

Throughout this document I use the words Comparison and Contrast interchangeably. For what we are doing, they mean the same thing. I thought that I ought to spell that out to avoid confusion.

$\endgroup$
  • $\begingroup$ In post-hoc comparisons following ANOVA, you either correct the test statistic (e.g., Tukey HSD) or the p-value (e.g., LS means or Bonferroni correction), to make the test less liberal. Further, repeated measures suggest a very specific design where you need to account for the matrix-covariance matrix. I believe this answer could be improved by discussing these two different issues. $\endgroup$ – chl Oct 31 '20 at 19:30

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.