contrast for within-subject variables
HI all,
I'm trying to understand how the contrast variables for within-subject factors are tested for signficance for frequentist ANOVA. A t-test is given, but the df are much higher than I expect (e.g., if i have an factor "distance" or "dist" with four factor steps running polynomial contrasts, and my sample is N = 24, I would expect df for the t-test to be 23; yet, they are 69). I come from SPSS, here they use F-Test (yet, the values should be tranformable by F = t^2). Yet, the results are not identical across both statistical programms (SPSS and JASP). Screenshots of the diverging results see below
I have searched the forum, the only post close to my issue is this one: https://forum.cogsci.nl/discussion/6674/contrasts-in-mixed-anova; yet, this topic is not really elaborated on.
Perhaps what would have would be to know on which R-package is the contrast analysis based on - I tried to find out, but did not find any explicit information about this.
I can elaborate more / upload examplatory data if needed (I'm not sure if I perhaps just did not find the right source, so therefore only this quick post).
Best
MerzS
Comments
I don't know the answer. But I predict that when an answer comes it will include phrases such as "R emmeans package," and "Welch-Satterthwaite degrees of freedom." ;-)
R
Hi Both,
This should help
Divergence between JASP and SPSS ouptuts when computing polynomial contrasts for within-sjts designs — Forum (cogsci.nl)
Thank you, that really helped!