JASP & normality residues

Comments

• EJ

  January 18

  Dear PSTO,

  I am not sure I understand the question...
• TarandeepKang

  February 27

  Hi Both,

  
  

  @EJ I think this might be a language issue, perhaps between French and English. I think the question is really about whether the Shapiro-Willik test is based on raw data or the residuals with regard to the t-test and correlation.

  Do I understand you correctly, @PSTO?

  Best,

  
  

  Tarandeep
• EJ

  March 15

  Ah, OK, I see the help file is not explicit about this. This is relevant for the one-sample t-test I think. I will ask our expert.

  EJ
• JohnnyB

  March 18

  Hi @PSTO ,

  In both tests, the Shapiro-Wilk test is on the residuals, because that is also what the normality assumption in these linear models is about. It is a common misconception that the variable itself ought to be normally distributed, since the normality of the data and the normality of the residuals are closely related. If the residuals are normally distributed, then the data will also be normally distributed, but the other way around does not necessarily hold, which is why it's better to test for normality in the residuals.

  I will update the helpfiles to make this more explicit.

  Kind regards,

  Johnny