[Bug]: Descriptives: Numeric ordinal variables give unexpected results

[larionowresearch-pixel]

JASP Version

0.95.0

Commit ID

No response

JASP Module

Descriptives

What analysis are you seeing the problem on?

Descriptive statistics

What OS are you seeing the problem on?

macOS Silicon

Bug Description

I am reporting a bug related to incorrect descriptive statistics values when using "ordinal" variables. I previously indicated the same or "technically very similar" bug in JASP 0.19.2, which was subsequently fixed in JASP 0.19.3 (see Descriptives: Numeric nominal variables gives unexpected results (440f765)). However, it seems this bug has appeared once again in JASP 0.95.0.

Please look at the two attached screenshots. On one screenshot, there are four variables (i.e., PHQ-4 items), with actual values from 0 to 3 in the data-set, which are correctly displayed (i.e., from 0 to 3) when I use "scale" variables. But if I change a type of variables from "scale" to "ordinal", JASP 0.95 incorrectly displays that my four variables (i.e., PHQ-4 items) have values from 1 to 4.

Expected Behaviour

Please correct this bug.

Steps to Reproduce

Open Descriptives, see min. and max. values when using a "scale" variable type. Then, change a type of variable from "scale" to "ordinal", and compare min. and max. values.

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.

[vandenman]

@larionowresearch-pixel thanks for opening this issue, I agree that something goes wrong here. What do you think of my suggestion below?

Ordinal variables are internally treated as 1, 2, 2, 4 with (in this case) the labels 0, 1, 2, 3. This means that, for instance, the mean is not defined (at least not meaningfully) and also not shown in JASP. However, some other quantities are defined, such as the mode. For the mode, we show the label of the most frequent value. For the minimum and maximum (and other quantities that can be meaningfully defined) I think we should similarly show the label of the quantity. This means that for an ordinal variable with labels zero < one < two < three we'd get the minimum (maximum) of zero and three respectively.

@larionowresearch-pixel would that align with your expectations?

@JohnnyDoorn any remarks?

[EJWagenmakers]

We have had this discussion before, right? (Will try to find the issue later). Should we be holier than the Pope wrt these measurement levels? This issue does need to be fixed prior to our hotfix release, I think.