[Bug]:Paired samples T-Test shows wrong sign

[bellhaeuser]

JASP Version

0.19.3

Commit ID

No response

JASP Module

T-Tests

What analysis are you seeing the problem on?

Paired Samples T-Test

What OS are you seeing the problem on?

macOS Silicon

Bug Description

In my data set, the T1-value is higher than the T2-value. When I insert the two variables into the left (T1) and right (T2) cell in the JASP interface, it calculates a positive t-value and a positive Cohen's d--although actually a decrease is happening which should give me a negative t and d.
This seems to be a new problem, at least I haven't seen this problem in the past.

T1T2simulation.csv

Expected Behaviour

When T1 is higher than T2, the t-value and the Cohen's d should be negative (--> decrease).
When T1 is lower than T2, the t-value and the Cohen's d should be positive (-->increase).

Steps to Reproduce

1. Open attached csv-file ("T1T2simulation.csv") in JASP.
2. Chose "Paired Samples T-Test".
3. Insert variable T1 first (into the left cell).
4. Insert variable T2 next (into the right cell).
5. Activate "Descriptive statistics" and "Cohen's d".

Log (if any)

No response

More Debug Information

No response

Final Checklist

• I have included a screenshot showcasing the issue, if possible.
• I have included a JASP file (zipped) or data file that causes the crash/bug, if applicable.
• I have accurately described the bug, and steps to reproduce it.

[tomtomme]

When T1 is higher than T2, the t-value and the Cohen's d should be negative (--> decrease).

Interesting.
I see your point regarding the perceived "decrease" and that it would be intuitive to see a negative value associated with it. This would also be consistend with correlation coefficients like pearsons r.
But the standard formula for the paired (and independent t-test) is sadly the other way round. It states: first value - second value. So if the first is higher, we get a positive result, despite the perceived decrease.
Should the formula be changed? Absolutely. But can we actually change it without creating even more confusion? I do not think so. This would need a coordinated effort that would change the formula at least in some textbooks and other stats software too... I guess. Also, we would need to change post hoc Tests of RM ANOVA too etc. They show the same behaviour.