Vincent Arel-Bundock

Oh yeah, totally. Thing is that this is not really a "mistake" per se. These are just targeting different estimands. The mistake is not being clear enough about our research question and the quantity that might answer it.

The syntax can be confusing at first, but the nice thing about having both options at finger tips is that it surfaces the (sometimes) important distinction between two quantities. Most authors used to print *something* without telling readers what actual quantity they reported.

Very nice talk. Thanks for sharing!

Exactly! The `newdata` argument behaves exactly as in base R; `variables` is a new shortcut for G-computation.

avg_predictions(variables="D"): Marginal means w/ counterfactual values of D.

avg_comparisons(variables="D"): ATE of D, i.e. difference between "counterfactual" marginal means

That's a really nice explanation from Julia! If you'd like the same basic info in different words (+ a Penguins example), see here: marginaleffects.com/chapters/com...

Yep, the logic is that this parallels the syntax in avg_comparisons(variables=). Both generate "counterfactual" predictions by duplicating the full dataset and modifying only the focal variable.

ok, yazi looks amazing. Had never heard of it before but I'm switching immediately.

#rstats
It is with profound sadness I heard that my long-time friend and colleague, John Fox passed away this week.
He was the author of {car}, {effects}, {Rcmdr}, ... and numerous influential books. I will miss him greatly.
www.john-fox.ca

Data Visualisation Gallery Gallery of data visualisations created by Nicola Rennie.

One of the things that has been on my to do list for a very long time, is building a gallery of all of the charts I've made across #TidyTuesday, #30DayChartChallenge, #30DayMapChallenge, and other miscellaneous projects 📊

And it's finally here!

Link: nrennie.rbind.io/viz-gallery/

#DataViz #RStats

Wow! Wish I'd thought of that before submitting 0.0.1 to cran!

classic bait and switch

i laughed

library (tidybayes) # for wrangling and plotting brms models Library (bayesplot) # for plotting brms models library (marginaleffects) #for marginal effects library (ggdist) library (patchwork)

Coding tip of the day: make sure you leave informative comments in your code

looks cool, congrats!

Right! And my impression is that the people who struggle most are those who insist on immediately computing some kind of "one number summary." I feel it's much easier to start simple, focusing on individual outcome levels, and then build up more complex contrasts piece-by-piece.

[as the maintainer of an R package that uses the world bank data API, I receive emails about this semi-regularly. it usually doesn't last too long.]

yep

this is super cool

#rstats Please welcome Jarl, a new R linter.

Jarl is a CLI tool with extensions in VS Code, Positron, and Zed. It can check thousands of lines of R code in milliseconds.

Jarl provides several output formats, a Github Actions workflow, and more.

Blog post: www.etiennebacher.com/posts/2025-1...

the stata thing is kind of wild. Congrats!

Our py-econometrics org has a new package: `maketables` is now installable from PyPi! You can create regression tables for PyFixest, statsmodels, linearmodels, and save them as html, latex, or word.

github: github.com/py-econometr...
docs: py-econometrics.github.io/maketables/

This is wonderful!

What's your cleaning/oiling routine? I must scrub mine too hard because it doesn't improve.

suuure. bsky.app/profile/did:...

Building an easy-to-use chrome extension to do this feels like a fun weekend project for a single motivated developer. Plus, there are already dozens of threads on Reddit asking for automated tools. Botnets are an obvious threat, but we wouldn't even need them to massively screw with our samples.

I also love this new form of social desirability bias, where the LLM infers the research question and gives researchers what they want.

new paper by Sean Westwood:

With current technology, it is impossible to tell whether survey respondents are real or bots. Among other things, makes it easy for bad actors to manipulate outcomes. No good news here for the future of online-based survey research

we're done

Great talk! This is very cool work and I look forward to trying the package! It feels like the evals will be highly dependent on the quality (and size) of the target column. And that's also likely to be the most expensive thing to collect. Have you seen some good discussion/guidance around that?

I am very happy that 𝗣𝘆𝗙𝗶𝘅𝗲𝘀𝘁 𝟬.𝟰𝟬.𝟬/𝟬.𝟰𝟬.𝟭 is now available on PyPi! The release includes a few technical changes and exciting new features =)