Wilcoxon W shows significant result, but shouldn't - help

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
PolarBears
Posts: 3
Joined: Fri Sep 10, 2021 11:29 am

Wilcoxon W shows significant result, but shouldn't - help

Post by PolarBears »

Hi!
I am using Jamovi for my Master's thesis. I am currently comparing the performance of a behaviour at two places in an enclosure. The thing is, that the result of the Wilcoxon W test that I use shows a significant result, even though it shouldn't, as my data set mainly consists of 0's. I have tried to put in another value than 0 (for instance replace all 0's with 1's), and still with a significant result.
I have made similar tests, also using Wilcoxon W, and with the same amount of 0's in the dataset, where the result came out as not significant.

How does Jamovi "treat" datasets with a lot of 0's? Does anyone have an idea of why the result is significant even though the final sample size (values higher than zero) shouldn't make it possible to finalize a test?

Thank you in advance! :)
Lærke
Attachments
Skærmbillede 2021-09-10 kl. 13.36.03.png
Skærmbillede 2021-09-10 kl. 13.36.03.png (230.67 KiB) Viewed 7302 times
User avatar
jonathon
Posts: 2613
Joined: Fri Jan 27, 2017 10:04 am

Re: Wilcoxon W shows significant result, but shouldn't - hel

Post by jonathon »

hey,

the implementation in jamovi makes use of wilcox.test() function in R, so you can have a noodle through its documentation:

https://www.rdocumentation.org/packages/stats/versions/3.6.2/topics/wilcox.test

wilcox.test(
x=c(135,0,0,0,120,0,0,100,0),
y=c(0,0,0,0,0,0,0,0,0),
paired=TRUE,
alternative='greater')

cheers

jonathon
User avatar
MAgojam
Posts: 421
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

Re: Wilcoxon W shows significant result, but shouldn't - hel

Post by MAgojam »

Hi, Lærke.

Looking at your screenshot, there is something wrong with the statistic, compared to the sheet data.
Seven degrees of freedom (instead of eight) are reported for Student's t and 5 pairs (instead of 6) of values were tied.
What version of jamovi are you using?

Cheers,
Maurizio
PolarBears
Posts: 3
Joined: Fri Sep 10, 2021 11:29 am

Re: Wilcoxon W shows significant result, but shouldn't - hel

Post by PolarBears »

Hi!

Thank you (both) so much for your input. I'm using JAMOVI 1.6.23.0.

Cheers,
Lærke
User avatar
jonathon
Posts: 2613
Joined: Fri Jan 27, 2017 10:04 am

Re: Wilcoxon W shows significant result, but shouldn't - hel

Post by jonathon »

this isn't really related to your query, but mz is right, there is something funny going on with the degrees of freedom and the number of pairs reported.

are you able to share the .omv file?

with thanks
PolarBears
Posts: 3
Joined: Fri Sep 10, 2021 11:29 am

Re: Wilcoxon W shows significant result, but shouldn't - hel

Post by PolarBears »

I have actually just tried to delete everything in the file and write the observations in manually (again). Now it shows eight degrees of freedom, six pairs tied and a p-value that is more realistic (or at least it is not significant now).
So the problem seems to be solved ':)
Post Reply