Hello,

I have three related questions:

1. Does conducting the Shapiro-Wilk Test on Residuals of a Within-Subject data set bring any valid insights on its normality? Considering that residuals from a within-subject design are not independent.

2. Also interpreting Q-Q and Residual Plots could be very subjective.. How does a strong deviation look like? i.e. at which point one could argue the plot shows a "strong" deviation from normality?

3. In case conducting Shaprio-Wilk Test is also valid, which model is more reliable? Linear mixed model (e.g. lmer) or Anova (e.g. aov_ez)?

Thanks & Regards,