Descriptives now converts ordinal to scale

[JohnnyDoorn]

As proposed here.

Benefits over current approach:

• users expect means for ordinal variables
  Benefit over old approach:
• Make treatment of ordinal as scale more salient to user (could even add footnote)

[JohnnyDoorn]

Tagging the R people for input on this

[FBartos]

assuming that no one actually wants to get descriptives for the orders of the ordinal variables -- which is most likely the case -- this is the most parsimonious solution!

[tomtomme]

Can the user still force this back to ordinal, if he really wants / needs to?

[EJWagenmakers]

Darn. This is a bit of a problem, right? What if the user really wants the results for the ordinal level?

[vandenman]

Wasn't the idea to have two input boxes though? One for ordinal/ nominal, and one for scale? Because this approach may lead to another list of issues:

• "Mode is wrong, it gives a value outside of the scale". User expects the most frequently occurred value for an ordinal variable, but it's converted to scale and then we use a density estimate.
• "Quantiles/ IQR incorrectly computed". User expects us to use the same quantile type as R, but we convert ordinal variables to scale and use the default type for scale data.

I'm also not entirely sure if information is lost when converting an ordinal variable to nominal in JASP.

[JorisGoosen]

Have you guys decided on whether to merge this or not?

As far as I can see the only thing we are now missing is "min" and "max" for the ordinal values.
But opening the labeleditor will show you that, cause the top one will be min, and the bottom one will be max...

(Which we could just show for nominals despite them not formally having such a thing. It would just be "lowest occuring label" and "highest occuring label" then)