Shapiro-Wilk different results

Discuss the jamovi platform, possible improvements, etc.
Post Reply
ravello
Posts: 15
Joined: Tue Oct 27, 2020 1:51 pm

Shapiro-Wilk different results

Post by ravello »

Hello
Responding to a subscriber of my channel about jamovi, I realized that when I calculate Shapiro-Wilk in 2 ways a) descriptive / normality and b) T-test / normality (moretest) I get different values, attached image.
greetings and thanks in advance for your answer
Attachments
jamovi shapiro wilk.png
jamovi shapiro wilk.png (31.4 KiB) Viewed 1487 times
decaux
Posts: 27
Joined: Wed Feb 19, 2020 1:35 pm

Re: Shapiro-Wilk different results

Post by decaux »

I think, in the first analysis you are looking at all the dependent variable data together, while in the second analysis, the S-W statistic takes into account the independent variable
ravello
Posts: 15
Joined: Tue Oct 27, 2020 1:51 pm

Re: Shapiro-Wilk different results

Post by ravello »

All the analysis is on the same variable, it will be then that the calculation of S-W changes when it is on an independent t-test.
This is the database used:
https://github.com/ravellom/DataScience ... t-test.omv
User avatar
MAgojam
Posts: 421
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

Re: Shapiro-Wilk different results

Post by MAgojam »

Hey @ravello,

The shapiro-wilk test is performed on the residues and not on the groups themselves, because putting all the residues together and testing is accepted practice.
This means that if the residuals of one group are not normally distributed, the residuals of all groups combined will also not be normally distributed.
Would you like to take a look here, at an old post, where Jth and I had given an example and given an answer.
https://forum.jamovi.org/viewtopic.php?f=6&t=1350&hilit=Shapiro#p4741

Ceers,
Maurizio
Post Reply