[Feature Request]: Allow filtering empty "scale" entries

[richlv]

Description

JASP allows to filter data in many different ways (https://jasp-stats.org/2018/06/27/how-to-filter-your-data-in-jasp/), and in some cases it's as easy as clicking on entries (like for nominal scale). Currently no "quick" filtering is available for "scale" data - filtering out records with empty fields could be useful.

Purpose

Easily filter out records that have missing "scale" data.

Use-case

A study has some participants dropping out. While they are kept in the original dataset, further analysis is to be performed with them excluded. This is relevant for analysis that does not directly involve the empty field.

Is your feature request related to a problem?

No response

Is your feature request related to a JASP module?

No response

Describe the solution you would like

When doubleclicking the header of a "scale" item, show to entries "empty" and "not empty" with checkmarks, similar to "Label Editor" for ordinal and nominal items.

Describe alternatives that you have considered

Custom filtering formulas, separate dataset.

Additional context

While sometimes this can be achieved by applying a custom filter ("item">0), this is not as discoverable. It would also be more tricky for items that can be negative.

[tomtomme]

Can't you use "Preferences => Data => Missing value list"?

Our new beta even has this functionality per variable in the "edit data" mode".
See: https://jasp-stats.org/2023/12/29/introducing-jasp-0-18-2-beta-process-survival-analysis-and-more/

But I may missunderstand the problem you want to solve.

[richlv]

"Missing value list" seems to be different, but maybe I'm not fully grasping it.
A more specific example:

• There's a dataset that has demographic data like age, and some metric.
• Metric is missing for some participants that dropped out.
• I'd like to perform some analysis on demographic data without modifying the data file.

[tomtomme]

I do not see why you would need to modify your data, if some cells are empty.
If some metric measurements are missing, JASP is calculating the analysis for the rest. You don't have to do anything. Other variables like age should not be affected by the missings in the other variables, except for multivariate analyses that needs all those vars (where you would then need to impute those missing values and that functionality is indeed missing in jasp)

About grasping the missing value list:
Consider the possibility, that you want to code missing values in a special way instead of leaving just an empty cell. E.g.:

• a participant does not want to answer a question could be coded via "-99"
• a measurement error because of a malfunctioning measurement device could be coded via "-88"
• etc.

those numbers are chosen in a way, that those could not occur under any circumstances. You then woul put those numbers (-88, -99) in the aforementioned "missing values list" so that jasp can recognize them as values, that should not be used when calculating the mean etc. But those values would be counted as missing e.g. in the frequency or descriptives table.

With this being said, can you now be more specific abaout what is missing for your usecase?

[richlv]

Sorry if I'm misunderstanding, will try to rephrase this.
Let's say I have a dataset with age, gender and some experimental metric like time.
Some participants filled in the demographic data, but dropped out of the experiment - thus they are present in the dataset, but the "time" metric is empty.
These participants should be excluded from analysis, but kept in the dataset. If some descriptive analysis is performed that does not involve "time" (like showing age split by gender), the invalid entries are included (because data for all participants contains the demographic values).
While I can add a filter like "time > 0" in this case to achieve the desired result, it would be great to have something easier and more generic - some easy way to filter out entries where one column is empty.

[tomtomme]

I think I get it now :D
But I do have no idea how this could be made easier than just setting the filter. Or even deleting those persons completely in data edit mode.

Do you have something in mind? Could you make a mockup of the way this should work and look?

[JorisGoosen]

Since 0.18.2 you can have empty values per column, also in scalar.

I think this would do what you want right?

[JorisGoosen]

Or you can just do !is.na(colName) no?

[richlv]

I think I get it now :D But I do have no idea how this could be made easier than just setting the filter. Or even deleting those persons completely in data edit mode.

Do you have something in mind? Could you make a mockup of the way this should work and look?

Deletion is pretty much keeping two files, if the raw data is to be preserved (if I understood this correctly).

As a user I wouldn't mind having something this simple ;)

The second thing I tried was the filter builder like so.

Tried to enter "" and NULL :)

Then I also tried this, but no luck either.

Since 0.18.2 you can have empty values per column, also in scalar.

I think this would do what you want right?

Sorry, not sure how this would work - is that the mapping for empty/NaN values? If so, it might not help in this case.

Or you can just do !is.na(colName) no?

Oh neat, I tried this in the R filter, and it does seem to work as desired.
Would it be feasible to expose it for less advanced users like me in some click-through way?

[JorisGoosen]

Oh neat, I tried this in the R filter, and it does seem to work as desired.
Would it be feasible to expose it for less advanced users like me in some click-through way?

Well, I think this could be a nice and easy feature to add.

We could add a "filter out missing values" checkbox to the "Missing values" tab of the columninfo editor.
This would then be for all columntypes.

[richlv]

That sounds excellent. Perhaps the wording or a tooltip could be "Filter out <rows|entries> with missing values in this column", to avoid potential confusion.

[JorisGoosen]

Well that is a bit too long to fit there but it can go into a tooltip for sure