df in JASP contrasts analysis
Hello,
I'm seeking clarity on the calculation of degrees of freedom for each contrast in a within-between interaction effect analysis in JASP. It seems the degrees of freedom vary between contrasts and differ from those in ANOVA, as well as from those that appear when I perform the same T-test separately to examine the same contrast. Could someone shed light on this, particularly regarding the occurrence of incomplete numbers in the calculations?
I would appreciate any help
Thank you,
Mor.
Comments
Hi Mor,
I think it would help if you show a concrete analysis with some screenshots...and have you read the background literature?
Cheers,
E.J.
Hi E.J.
I would like to pick up on the previous question. I am also trying to understand how exactly the degrees of freedoms for planned contrasts are computed. I understand the the R package emmeans is used and I suspect that the difference between a regular paired-sample t-test and using planned contrasts has to do with the planned contrast using the pooled error term from the ANOVA and the fact that the underlying emmeans package will compute the denominator degrees of freedom based on the estimated marginal means from the full ANOVA model.
Can you verify that this is conceptually correct? Is there a formula that will help me understand the underlying computation that results in df = 111.5 in my case rather than df = n - 1 = 24 with a paired-sample t-test. Below I provide two screenshots, where the highlighted line refers to the contrast between the same condition pair (Odor-Left vs. Odor-Both in the second sequence half).
The corresponding rmANOVA model is a 4 x 2 design with 25 participants and I tick the box 'use pooled error term for follow-up tests'. I would be happy to provide additional information if needed.
Thanks! Kind regards,
L.