df in post-hoc tests

[zf383962]

I was hoping someone might be able to provide me with some guidance regarding some weird df I’ve been getting in post-hoc tests following up a repeated-measures ANOVA. For some context, the study I’ve run is fully within-subjects, with 2 variables that each have 4 levels- a condition variable (condition A, B, C, D) and a secondary variable (levels 1, 2, 3, 4). We collected data from 300 subjects. I ran a RM ANOVA and also ran Bonferonni corrected post-hoc t-tests comparing the 4 conditions collapsed on the secondary variable (i.e. condition A vs. B ), and then comparing within each condition based on the secondary variable (i.e. within condition A, comparing level 1 vs. 2). Jamovi gave me 897 df for the condition comparisons collapsed on the secondary variable, and 3,586 df for the within-condition comparisons of the secondary variables.
In trying to sort out how this happened, I found this thread on the forum (viewtopic.php?f=2&t=741) where someone said that for RM ANVOA, Jamovi uses the full Welch method including the Welch-Satterthwaite equation for df. But when I plugged in my sample size and variance into the equation on the page they linked to (https://en.wikipedia.org/wiki/Welch%27s_t-test), it gave me a very small value that was nowhere near what Jamovi was giving me. Anyone have any idea how these df were determined? Any help or guidance would be really appreciated! Thanks.

[jonathon]

hi,

the df's are provided by the emmeans package. here's the code:

https://github.com/jamovi/jmv/blob/master/R/anovarm.b.R#L590-L630

i'm personally not that clear on all this stuff, but my understanding is the emmeans way is the *correct* way.

cheers

jonathon