I need to compare two groups of participants to each other in a model that contains fixed and random effects in R. It was suggested to me that a non-parametric test might be ideal for these comparisons. Here is the data:
structure(list(item = c("'accordion_1'", "'accordion_3'", "'apple_01'",
"'apple_03'", "'asian_01'", "'asian_02'"), avg = c(3.177631579,
2.868421053, 4.421052632, 4.526315789, 4.440789474, 4.0625),
subcat = c("c", "c", "g", "g", "h", "h"), cat = c(1L, 1L,
2L, 2L, 2L, 2L), group_num = c(1L, 1L, 1L, 1L, 1L, 1L), group_name = c("adu",
"adu", "adu", "adu", "adu", "adu")), row.names = c(NA, 6L
), class = "data.frame")
I have scoured the internet for code to accomplish this (I am new to both these methods). I initially started by using perm.lmer() in the permutes package with this code:
asd_kid_mod <- perm.lmer(avg ~ group_name * cat + (1|item),
data = asd_kid, nperm = 1000, type = "anova")
but it returned p values at 0 which seems strange to me. I also am not sure if mean permutation testing would simply be more appropriate for my data. So I guess my questions are:
- What is the best method for permutation testing for LMMs in R?
- Would it make more sense to use mean permutation testing?
EDIT:
Here is my output form summary:
Factor df LRT F p
1 (Intercept) 1 555.54630 253.445582 NA
2 group_name 1 29.48206 1.643257 0
3 cat 1 220.76503 24.153896 0
4 group_name:cat 1 64.43601 4.027463 0
and here's a link to what I mean by mean permutation testing: https://towardsdatascience.com/how-to-use-permutation-tests-bacc79f45749
I believe it would be using means instead of medians based on the link above
