Test of Normality suddenly not significant?

General help and assistance with jamovi. Bug reports can be made at our issues page: https://github.com/jamovi/jamovi/issues . (If you're unsure feel free to discuss it here)
Post Reply
Queezle
Posts: 2
Joined: Tue May 17, 2022 2:19 pm

Test of Normality suddenly not significant?

Post by Queezle »

Hi,

so I want to compare two dependent variables that do not meet the criteria for t-test because Shapiro-Wilk test is sigificant in both cases. I ran a t-test nonetheless in order to be able to get Wilcoxon W. When ticking the "normality test" box, jamovi suddenly showed me a non-significant result (it looks like both variables were somehow merged because they are not shown individually). Can someone explain me how this is possible?

Kind regards.
User avatar
jonathon
Posts: 2613
Joined: Fri Jan 27, 2017 10:04 am

Re: Test of Normality suddenly not significant?

Post by jonathon »

hi,

provide us with a data set, and the steps to reproduce what you describe, and we'll be able to provide some insight.

cheers

jonathon
Bobafett
Posts: 76
Joined: Thu Jul 18, 2019 11:33 am

Re: Test of Normality suddenly not significant?

Post by Bobafett »

This is because you ran a Shapiro-Wilk on both variables. A paired t-test is specifically interested in the difference in each pair of scores, and so the singular (non-significant) Shapiro-Wilk is actually testing the distribution of these differences - which is an assumption of this particular paired t-test.

To demonstrate this - create a new computed variable based on a simple 'Variable1 - Variable2' calculation. This will give you the difference in the pair of scores for each participant - then you can run a Shapiro-Wilk on this new 'differences' variable. It should give you the same SW and p-value you found when doing this as part of the paired t-test options!
Queezle
Posts: 2
Joined: Tue May 17, 2022 2:19 pm

Re: Test of Normality suddenly not significant?

Post by Queezle »

Bobafett wrote:This is because you ran a Shapiro-Wilk on both variables. A paired t-test is specifically interested in the difference in each pair of scores, and so the singular (non-significant) Shapiro-Wilk is actually testing the distribution of these differences - which is an assumption of this particular paired t-test.

To demonstrate this - create a new computed variable based on a simple 'Variable1 - Variable2' calculation. This will give you the difference in the pair of scores for each participant - then you can run a Shapiro-Wilk on this new 'differences' variable. It should give you the same SW and p-value you found when doing this as part of the paired t-test options!
Thank you very much! Fortunately, I found out the answer shortly after posting - but I still very much appreciate your help. Have a nice day!
Post Reply