Question

我进行了重复测量的方差分析，其中的相互作用效应几乎是显着的（F = 3.96，p = 0.053）。但是，比较相互作用模型和主效应模型的贝叶斯因子偏爱主效应模型（BF〜0.3）。考虑到方差分析中相互作用的作用接近显着性，BF是否应该更接近1？

我已在下面复制了我的R代码，可以在here中找到数据。

library(dplyr)
library(BayesFactor)

my_df <- read.table("~/Downloads/mm_data.txt", header = T, sep = '\t')

aggdf <- my_df %>% 
  group_by(ID,var1,var2) %>% 
  summarize(outcome = mean(ACC, na.rm = T)) %>% 
  ungroup() %>% 
  mutate(ID = as.factor(ID), var1 = as.factor(var1), var2 = as.factor(var2))
aggdf <- as.data.frame(aggdf)

ezANOVA(aggdf, dv=outcome, wid=ID, within=var1*var2, type=3)
# Note that the interaction effect is almost significant (p=.053)

bf_all <- lmBF(outcome ~ var1 + var2 + var1:var2 + ID, data = aggdf, whichRandom = "ID", progress = FALSE)
bf_me <- lmBF(outcome ~ var1 + var2 + ID, data = aggdf, whichRandom = "ID", progress = FALSE)

#BF10
bf_all/bf_me # this should be closer to 1 (or >1)

R BayesFactor计算与对象内部方差分析的结果不一致

0 个答案: