[Feature Request]: Printing Standardized Residuals (ANOVA)

[RachelKing42]

Description

Save the standardized residuals as a variable

Purpose

to check normality

Use-case

No response

Is your feature request related to a problem?

No response

Is your feature request related to a JASP module?

ANOVA

Describe the solution you would like

I would like to be able to save the standardized residuals in order to check normality using methods other than the QQ plots

Describe alternatives that you have considered

The ability to check skew/kurtosis and look at histograms/boxplots on the standardized residuals in the ANOVA tab rather than printing them. Also a method to flag outliers that would enable easy location and filtering of cases.

Additional context

I was teaching ANOVA on SPSS and used to get the students to save the standardized residuals and then check Z-Scores for skew and kurtosis, histograms and boxplots. We also used to use the standardized residuals to check outliers. Although the QQ plots are available I would like them to use a couple of different methods to check normality. I can see there is the option to save the residuals in the regression tab and I am hoping for the same functionality in the ANOVA tab (they have not learnt regression yet). Many thanks!

[tomtomme]

@RachelKing42
Thx for the certainly valid request.
Until this is implemented (which might take some time) you already can do some of the things you requested - kind of. Have you tried the Raincloud Plots within the ANOVA analysis? This could highlight outliers and departures from normality per group on the raw data scale - so not for standardized residuals as you request. But I hope it may help you as a workaround in the mean time.
Cheers, Thomas

[tomtomme]

This is related to: #2878

[RachelKing42]

@tomtomme

Many thanks for your speedy reply. I do really like the raincloud plots!

All the best,
Rachel

[JohnnyDoorn]

Hi @RachelKing42 ,

Yes, this is an excellent suggestion and is very high on my priority list for when I start working on JASP again after my teaching period, though that will only be in January

Cheers,
Johnny

[RachelKing42]

Hello Johnny and Thomas,

Many thanks for your reply and excellent to hear that the functionality will be available in the future!

For now I am wondering whether I can calculate the residuals in JASP using the R interface. To do so I need to be able to subtract the group mean, rather than the grand mean .. so I am looking for a method of splitting the data that would work in JASP when creating a new variable. Is this something that would be possible? I have been playing around without luck so thought it might be best to ask rather than continuing down a rabbit hole. :)

I have considered excel, but I think the approach might be a little too finicky for this level of study when it comes to the factorial designs. So if it is not possible within JASP I will probably stick to checking the raw data this time around.

Many thanks!
Rachel

[JohnnyDoorn]

Hi Rachel.

You can add a new column with the group means using tapply in R-mode:
tapply(Heart Rate, Group, mean)[Group]
This will have the group means for each participant. You can then create another column that takes the difference between the new column and the grand mean (or do it directly in the first computed column).

However, I also fail to do it for factorial designs, but what you can do is do the ANOVA as a regression, export the residuals there, and then have a column defined as "DV - residuals", which then does give you the group means for the full factorial design (if you include the interaction in regression). I've attached the jasp file HeartRatePredictions.zip (change extension from .zip to .jasp). Hopefully that helps (in any case this was a fun puzzle!)

Cheers,
Johnny

[RachelKing42]

Awesome! Thank you! :D