Robust mixed Anova in R - Interpretation, post hoc tests and effect size

368 views Asked by At

I tried to perform a robust mixed anova in R because the assumptions of homoscedasticity, normal distribution, and equality of covariances were violated. I have one between subject factor (group) and one within subject factor (situation). I have two questions in total.

Question 1: How do I get the corresponding effect sizes?

mixedanova <- WRS2::bwtrim(formula = value ~ group*situation, id = id, data = data.long)

Output:

value df1 df2 p-value
group 162.9427 2 33.9744 0
situation 136.4997 1 38.5618 0
group:situation 61.4333 2 28.2040 0

Question 2: How do I interpret the post hoc tests? (Why are there values greater than 1?) (Why does the output not change when I change nboot?)

sppba(formula = value ~ group*situation, id = id, data = data.long, est = "mom", nboot = 500)

Test statistics:

Test whether the corrresponding population parameters are the same:
p-value: 0
group Estimate<dbl>
1-2 0.8178253
1-3 1.4685165
2-3 0.6506912
sppbb(formula = value ~ group * situation, id = id, data = data.long, est = "mom", nboot = 500)

Test statistics:
Test whether the corrresponding population parameters are the same:
p-value: 0 
Estimate<dbl>
situation 1-situation 2 0.2575347
sppbi(formula = value ~ group * situation, id = id, data = data.long, est = "mom", nboot = 500)

Test statistics:
Test whether the corrresponding population parameters are the same:
p-value: 0 
header 1 header 2
situation1-situation2 1-2 -0.58068783
situation1-situation2 1-3 0.07747543
situation1-situation2 2-3 0.65816327

Thanks for your help in advance!!

I have read through the paper by Mair & Wilcox (2020) "Robust Statistical Methods Using WRS2" and have tried varying the functions bwtrim, sppba, sppbb, and sppbi, but unfortunately without success.

0

There are 0 answers