Phi Fei Fo Fum

[mattansb]

This PR is for writing the Phi Fei (V, W, T) paper, potentially for https://github.com/orgs/easystats/teams/core-team/discussions/5

[mattansb]

Paper is currently a first draft, I invite any one willing to participate to lend a hand in editing / writing / formatting (don't mind switching to qmd instead or rmd, or even MSWord tbh...) @easystats/core-team

[codecov-commenter]

Codecov Report

Merging #548 (180d8cf) into main (bb38e03) will not change coverage.
The diff coverage is n/a.

❗ Current head 180d8cf differs from pull request most recent head b775fc7. Consider uploading reports for the commit b775fc7 to get more accurate results

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

@@           Coverage Diff           @@
##             main     #548   +/-   ##
=======================================
  Coverage   89.56%   89.56%           
=======================================
  Files          56       56           
  Lines        3391     3391           
=======================================
  Hits         3037     3037           
  Misses        354      354           

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[IndrajeetPatil]

Thanks, @mattansb, for getting the ball rolling! Looks like a great start.

For now, I think we should work off of an .Rmd doc, which supports a lot more article formats via the {rticles} package.

I have already added some commits, and will keep editing as I find time.

Btw, currently, fei is not printed in the PDF output. I found this solution to fix this. Need to try it out.

[IndrajeetPatil]

Team, please include [skip ci] in your commit messages and avoid unnecessary energy wastage through CI/CD, which is not relevant for this PR, at least as long as it remains laser-focused on the paper.

[rempsyc]

I started working on the paper a bit and made some changes. I had [skip ci] in my working memory (and memory pad) right up until I committed the message 🙈 so cancelled all corresponding GHAs after they started. For now, I've fixed the blank פ issue by using unicode. I don't know if having it as a math-mode symbol has a distinct advantage for us here though.

[mattansb]

Thanks Remi (and ChatGPT)! Please add yourself to the paper (:

[rempsyc]

Ok! @mattansb what do you think of this new abstract/intro/conclusion?

[mattansb]

Looks good to me! Thanks Remi!

[strengejacke]

I looked at the current draft of the paper - do you think it will be realistic to submit it to the special issue? There's still some work to be done, in particular some kind of methods section, or at least some separation of the "theoretical" part and "results" part, and the discussion. Futhermore, I think we should use proper tables or in-text reporting of results, and exclude the R code (this is for supplemental).

[bwiernik]

A few revisions and comments on sections we should add for clarity. Looking good!

[bwiernik]

I think we need a short section here on effect sizes in general and describing Cohen's d and Pearson r and why they are inadequate for categorical data. Also commenting on odds/risk ratios and why we might not like them (hard to interpret, not bounded 0-1). Basically say that we want effect sizes appropriate for categorical data that are in a correlation like metric.

[mattansb]

Something like this?

• Odds ratios and risk ratios are only good for 2x2 table
  • RR is also not symmetrical (columns and rows are not exchangeable)
• Pearson's r is appropriate for two qualitative variables (cf. biserial correlation) for which (co)variance is meaningful.
• Cohen's d is applicable for a qualitative outcome (and is not symmetrical) for which a mean and variance are meaningful.

What else?

[bwiernik]

We should add a short section to each part above demonstrating not only the tables that produce 0 and 1 but also some "small", "medium", and "large" correlations to help readers get a sense of what different values of these effect sizes look like in data.

[mattansb]

Not sure what you mean..?

[bwiernik]

in addition to the real-data example, we should also demonstrate, eg, what different "benchmark" small, medium, and large correlation values (e.g., .1, .3, .5, .7, .9) look like in a contingency table

[bwiernik]

@rempsyc did those "new statistics" changes make it into the Rmd?

[rempsyc]

No, as we have abandoned the Rmd to work exclusively on the word doc, in my understanding.