Vincent Arel-Bundock

<-- Still using vim+tmux 90% of time in 2025.

yesssss!

too kind

This workshop is going to be great fun, and I really look forward to meeting everyone!

It would mean a lot to me if you signed up!

Glad you like it!

lol

Today, I created a massive 1000 page-long table.

It only took FIVE SECONDS to render the PDF using the tinytable 📦 for #Rstats, @quarto.org and @typst.app !

v0.11.0 of tinytable speeds up Typst support, and brings feature parity with HTML and LaTeX

Docs: vincentarelbundock.github.io/tinytable/

That 37% stat is wild.

One of the best polisci departments in Canada. The people there are great. Apply or tell your students!

Mostly because people look down on the activity, and I hope that a fancy label will help.

I just learned the expression "translational methodology" and I think I like it.

This is grim reporting indeed, but I have to admit that I enjoyed this clip.

"It's not easy to look tough with a flute, even if it is German. [...] But the audience is rapt."

news.sky.com/story/inside...

Oh, that's a nice way of putting! At times it was pretty clear that something was not going to work, and for some reason I kept arguing instead of doing it myself.

Sometimes, it does really feel magical. It found some obscure bug in 1 minute that would have taken an hour for me to identify. Great of menial tasks. But then it sent me deep into wild rabbit holes. What a weird tech!

It also kept ignoring instructions in CLAUDE. md file, so I'm not sure how to enforce a "never change tests" instruction.

lol. Just spent 100$ in a single day for a stupid vibe coding hobby/experiment. I don't even feel it was an intense coding day... I can't afford this tool.

Also, at some point Claude made changes to the code that broke some tests. They tried to change the tests instead of fixing the code.

I may be more tool-agnostic than you are. These are all fine and pretty much equivalent. I just have personal preferences, and since most of what I work on is hobby-coding or personal research projects, I'm free to have opinions. But I'd be fine (and easy) to learn any of them for money if I had to.

I mean it's fine to teach "data manipulation in X" as long as X is sufficiently close to Y that the general concepts apply and the students can trivially transfer.

I think I disagree with this. It's easy to pick up pandas when you know Polars. It's fine to teach transferable skills rather than focus on a specific tech stack.

If you still have energy, you might try Polars. I mostly quit python years ago because I couldn't wrap my head around Pandas, but have found Polars fantastic. YMMV, of course.

FYI, I wrote a python version of marginaleffects that works with both of those. It's far less mature and still has many bugs, but it should basically works.

In our polarized political environment, I’m excited to share my new ASR publication: “Competence over Partisanship: Party Affiliation Does Not Affect the Selection of School District Superintendents”. journals.sagepub.com/doi/10.1177/...

There have been a number of recent articles on statistical power in quantitative political science. This is something that I think deserves more attention and discussion. A short thread of the articles I have read. 🧵

Group Bootstraps — group_bootstraps Group bootstrapping creates splits of the data based on some grouping variable (which may have more than a single row associated with it). A common use of this kind of resampling is when you have repe...

Ah right! There's also `group_bootstraps()`. Maybe that helps? In principle, it should be trivial to support that in `marginaleffects::inferences()`. Just look for a `group` argument in `...` and swap out the resampling function. rsample.tidymodels.org/reference/gr...

In principle, this should be possible because `rsample::bootstraps()` accepts a `strata` argument, and `inferences()` passes all `...` to that function.

Within-person factorial experiments, log(normal) reaction-time data | A. Solomon Kurz Causal inference with the GLMM, Part 1

New #rstats blog up!

solomonkurz.netlify.app/blog/2025-07...

This is the first in a new series discussing causal inference with experimental data using multilevel models. My basic case is g-computation is the way to go.

Section 6.3.1 of the book explains the difference. It's a bit subtle. marginaleffects.com/chapters/com...

yessss