Incomplete labeling of outliers

[karenzagorski]

• JASP version: 0.10
• OS name and version:Windows, Version 10.0.17134 Build 17134
• Analysis: Side-by-Side boxplots with labeled outliers
• Bug description: Some outliers don't get labels
• Add JASP files as a zip:bug report.zip1
• Screenshot:
• Expected behaviour: All outliers should have a label
• Steps to reproduce:

1. Descriptive statistics. Lifespan(variables) vs exposure(split)
2. Plot as box plot. Check label outliers
3. Groups 2 4 and 5 have unlabeled outliers, group 1 is labeled properly

[TimKDJ]

@AlexanderLyNL in your PR jasp-stats/jasp-desktop#3320 you changed the quantile algorithm type from 7 (R standard) to 6 (SPSS standard). This gives different quantiles. ggplot does not allow you to set the quantile type and by default uses type 7. I also noticed we use quantile() in other analyses and do not use type 6.

I think the only way to change the ggplot is to completely overwrite the boxplot element and add our own outlier dots. Should we really perform this switch from 7 to 6?

[EJWagenmakers]

I have a strong preference for whatever R uses over whatever SPSS uses...

[AlexanderLyNL]

Sure we can change everything from 6 to 7. Would you like to do this? Or should I?

[TimKDJ]

Yeah I can do it -- I was just wondering what the best course of action was

[AlexanderLyNL]

I thought that it would be good to have the output consistent with SPSS, but that doesn't seem to be the case, so let's follow the R default instead.