Correlation data different in SPSS and Jamovi

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)

by lucier96 » Wed Dec 09, 2020 5:18 pm

I have calculated various Spearman's correlations using SPSS. However, I want to conduct a partial correlation but SPSS can only do this for Pearson's. After playing around with Jamovi, I noticed that both the test statistics and p values are different on SPSS and Jamovi for the original correlations (selected one-tailed on SPSS and selected 'correlated' option rather than 'correlated positively' or 'correlated negatively' on Jamovi). The data is exactly the same as I loaded the SPSS document into Jamovi.

I then compared the data between SPSS and Jamovi for partial correlations (Pearson's) and this is also different.

Does anyone know why this is happening and how it can be solved? I am fairly new to Jamovi so apologies if it is something obvious.
lucier96
 
Posts: 2
Joined: Wed Dec 09, 2020 5:11 pm

by jonathon » Wed Dec 09, 2020 9:25 pm

so i wouldn't assume there's anything to be "solved" here.

there's a few possibilities; jamovi does this incorrectly, spss does this incorrectly, jamovi and spss use slightly different methods, your analyses between jamovi and spss are not set up *exactly* the same, etc.

double-check that the N's are the same between spss and jamovi, and if that's not the issue, then feel free to post your data file here.

(back in the early days, when people posted "the result isn't the same as spss", there was an even chance that our implementation isn't quite correct. but these days, everything's been so thoroughly tested, it's very rare that someone finds a value that isn't correct. the last time that happened was probably two years ago).

cheers

jonathon
User avatar
jonathon
 
Posts: 1550
Joined: Fri Jan 27, 2017 10:04 am

by Bobafett » Thu Dec 10, 2020 11:44 am

selected one-tailed on SPSS and selected 'correlated' option rather than 'correlated positively' or 'correlated negatively' on Jamovi

Wouldn't this mean you requested a one-tailed correlation in SPSS but two-tailed in jamovi?

Not sure about the partial correlation - like Jonathon said, you'd need to post screenshots or the data file to get some help...
Bobafett
 
Posts: 13
Joined: Thu Jul 18, 2019 11:33 am

by MAgojam » Fri Dec 11, 2020 1:02 am

Hi, @ lucier96.

If a one-tailed Pearson correlation is selected in SPSS, the p-value may differ from that reported by jamovi, only because SPSS always considers a correlation to be positive, but when it is negative it returns the one's complement of the calculated p-value. Jamovi always looks at only one tail, right or left depending on whether you have chosen positively or negatively correlated.
Try choosing 'positively correlated' for one-tailed in jamovi and one's complement the p-value of a negative correlation, you will see that it will be the same as that reported for one-tailed by SPSS.

Cheers,
Maurizio
User avatar
MAgojam
 
Posts: 125
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

by lucier96 » Fri Dec 11, 2020 2:41 pm

Thanks for all the useful suggestions.

I have selected 'correlated positively' in Jamovi and this is now the same as my SPSS results for the majority of my correlations. For the correlations which have come out negative, do I select 'correlated negatively'? as these are the p-values which seem to differ from SPSS.

Thanks!
lucier96
 
Posts: 2
Joined: Wed Dec 09, 2020 5:11 pm


Return to Help