Roadmap for release 0.6.0

[raphaelvallat]

The following issues should be addressed:

• anderson() boolean returned value is inverted #308 ❗
• Feature request: add confidence interval to pg.wilcoxon() #153
• Adding eta squared parametric ci to pingouin.compute_esci #253
• Naming of return-values for T-tests etc #208
• Return Confidence Interval for nonparametric Mann Whitney U Test #225
• Get residuals from anova #118

The following PR should be merged:

• Add confidence interval for MWU #226
• Add new function for pairwise T-tests between columns of a dataframe (pingouin.ptests) #291

The following dependencies should be updated:

• Scikit-learn <1.1.2 (revert temporary fix in Use scikit-learn<1.1.0 #278)

The following deprecation should be made:

[Federico2111]

Seems like a good new release. Is there any time line? When will these upgrades be made?
Thanks

[raphaelvallat]

Hi @Federico2111,

I have had no time to work on Pingouin in the last few months, and I anticipate this will continue at least until Q1 2023. I therefore cannot promise a timeline for this new release.

[musicinmybrain]

The following dependencies should be updated:

* [ ]  Scikit-learn <1.1.2 (revert temporary fix in [Use scikit-learn<1.1.0 #278](https://github.com/raphaelvallat/pingouin/pull/278))

This can be checked off, since #300 is merged and released in 0.5.3.

[Federico2111]

Any update about the progress of this new release? Any help needed?

[raphaelvallat]

Thanks @Federico2111 — unfortunately at the moment I only have time for urgent bug fixes, not for implementing new features. So it's unlikely that version 0.6.0 is coming anytime soon; however I'd like to release version 0.5.4 with some minor improvements and bugfixes (e.g. CI tests) in the coming weeks.

[Federico2111]

Thanks @raphaelvallat.
By CI tests, you mean also the parametric Confidence Interval around eta squared that I had suggested?
#253

[raphaelvallat]

No sorry I meant the unit tests (aka continuous integration or CI), see GitHub Actions: https://github.com/raphaelvallat/pingouin/actions

[Federico2111]

Any update? :-)

[raphaelvallat]

I recently released a minor version that included bugfixes and some improvements: https://github.com/raphaelvallat/pingouin/releases/tag/v0.5.4

I cannot say when the next release will be, but likely not in the coming weeks. Feel free to submit PRs for what you consider to be the most important/urgent tasks. Thanks

[FlorinAndrei]

@raphaelvallat What's happening with the site https://pingouin-stats.org/ ? I see the text, but all the styling is broken. I've tried Shift-reload, no luck. Using latest Chrome on Win10.

[raphaelvallat]

@FlorinAndrei #400, I'll investigate and try to fix soon

[Federico2111]

Thanks @raphaelvallat
My suggestion was #253 which is to add the parametric confidence interval around the eta squared effect size. I provided some documentation about it. This feature would complete the pingouin.compute_esci library, where parametric ci's are calculated for other effect sizes, but not for eta squared: https://pingouin-stats.org/build/html/generated/pingouin.compute_esci.html

[turkalpmd]

I have implemented some of the adjustment methods mentioned in this article (https://scholarworks.umass.edu/pare/vol20/iss1/8/) in Python. If deemed appropriate, I can create a PR.

[raphaelvallat]

I have implemented some of the adjustment methods mentioned in this article (https://scholarworks.umass.edu/pare/vol20/iss1/8/) in Python. If deemed appropriate, I can create a PR.

@turkalpmd Can you please first open an issue to describe exactly what new or updated methods could be implemented based on this paper?

[raphaelvallat]

@Federico2111 I'm trying to move away from a system in which I implement external requests for features, towards a more sustainable model where people actually implement and submit modifications in pull requests. Furthermore, this is not a highly-requested feature so I have not considered it a high-priority item. Have you made your own implementation of this feature with Python?

[turkalpmd]

I have implemented some of the adjustment methods mentioned in this article (https://scholarworks.umass.edu/pare/vol20/iss1/8/) in Python. If deemed appropriate, I can create a PR.

@turkalpmd Can you please first open an issue to describe exactly what new or updated methods could be implemented based on this paper?

Here, I've opened the issue you mentioned, @raphaelvallat

[Federico2111]

Hi @raphaelvallat,
I have not implemented that feature (#253) in my Python scripts. If it is not implemented in the pingouin library, I would consider studying the literature and writing the code in my scripts. Currently pingouin.compute_esci calculates the parametric confidence intervals around the most commonly used effect sizes in one-way anova. Eta squared is the most commonly used effect size in two-way anova, which is why I think it would complete the features of the library and would be a beneficial addition.