Max kuhn

orbital 0.5.0 orbital 0.5.0 is on CRAN! More models and faster execution.

I'm over the moon excited to the release of 0.5.0 of orbital 🛰️

This release adds full support for boosted tree, faster creation of orbital objects, optimization of execution!

We can finally reliably predict with a xgboost model from a database!

tidyverse.org/blog/2026/03...
#rstats #tidymodels

something big is coming 🛰️

I’ve also made some headway on the python companion, but I’m not quite ready to release that yet.

Slow and steady! I’m hoping to have (single) classification tree chapter done in the next month and then probably off to boosting.

It’s hard to determine what a release should be since we’re going nonlinearly through the chapters (that pun was intended).

🤩

David Robinson - Teach the Tidyverse to Beginners YouTube video by Lander Analytics

@drob.bsky.social’s talk from 7 years ago covers this well

youtu.be/dT5A0sAWc2I

I presented at Shiny in Production 2025, an incredible conference hosted by @jumpingrivers.com up in Newcastle! I was glad to share the very latest from the Shiny team directly. The topics were bleeding edge at the time, so still new and relevant now. My video: www.youtube.com/watch?v=vxai...

We are covering 40 people's travel, lodging, and registration for posit::conf() this fall! If you are from a group that is underrepresented in data science or open source, please consider applying for the Opportunity Scholarship—we'd love to have you join.

posit.co/blog/apply-t...

March's tabular playground
#rstats #databs #tidytuesday
www.kaggle.com/code/jimgrum...

#tidymodels now has its very first cheatsheet! "Preprocessing data with {recipes}" is now available in Web and PDF versions here: rstudio.github.io/cheatsheets/... #rstats #posit #rstudio

A very helpful post on AI helpers, especially if you are new to AI and Claude

SAS’s sums of squares “types” was the ancestral language war with S(plus) back in the day. So quaint compared to now.

Too bad Claude can’t shovel snow.

Similarly, clinical trial people started calling observational data or non-randomized trials "real-world data" like _that_ is the abnormal case.

Yeah, it basically means a model built on the standard rectangular data structure.

That's what most of the world's data is, but since deep learning is all about images and non-tabular data (e.g., text), we have to give it a special name like it's the exception.

Here’s a clip from Max Kuhn (@topepo.bsky.social) of Posit breaking down how we can truly quantify LLM performance using a clear, generalizable framework.

See the full conference talk here: youtu.be/TQKbaIR-8J4

#AI #MachineLearning #DataBS

How to choose the best LLM using R and vitals Use the vitals package with ellmer to evaluate and compare the accuracy of LLMs, including writing evals to test local models.

Want to check if code using #GenAI generates the responses you want? Here's how to automate LLM evals with the {vitals} #RStats 📦 by @simonpcouch.com @posit.co
My latest at #InfoWorld:
www.infoworld.com/article/4130...
#LLMs

random.seed(42)

</cringe>

We owe sooo much to @yihui.org Thank you!

Let me put it this way… it was much much easier than trying to convince the same AI to just stop using the test set over and over and over again. 😃

With bookdown.org being decommissioned, I've been working on converting #rstats books from #bookdown to #quartopub. After spending a lot of time with Claude to get it right for my two books, here's a repo that might help you do the same (especially with Claude Code):

github.com/topepo/bookd...

I’m like this with tikz (tikz.net). The best looking figures and diagrams but I don’t have the time to sort through its arcana.

Dumping the main points of both these posts into a skill is a really easy way to use claude to try the conversions (and check unit tests) to see if these changes can help speed things up without regressions.

A while back, @simonpcouch.com wrote this relevant post for package maintainers to help them convert code from dplyr/tidyr to vctrs.

tidyverse.org/blog/2023/04...

`dplyr::if_else()` and `dplyr::case_when()` are up to 30x faster dplyr 1.2.0 comes with much faster and more memory efficient `if_else()` and `case_when()` functions!

Last week we released dplyr 1.2.0, but we left off something VERY important 🙂

`dplyr::if_else()` and `dplyr::case_when()` are now up to 30x faster and use 10x less memory!

We dive into how we achieved these numbers in this new #rstats post!

tidyverse.org/blog/2026/02...

For more than a year I have been working on a brand new Jupyter Notebook editor for Positron. This is a ground-up build of a new Jupyter Notebook experience built to leverage all the knowledge and tools Posit/Positron brings to the data science table. 🧵#jupyter

dplyr 1.2.0 dplyr 1.2.0 fills in some important gaps in dplyr's API: we've added a new complement to `filter()` focused on dropping rows, and we've expanded the `case_when()` family with three new recoding and re...

dplyr 1.2.0 is out now and we are SO excited!

- `filter_out()` for dropping rows

- `recode_values()`, `replace_values()`, and `replace_when()` that join `case_when()` as a complete family of recoding/replacing tools

These are huge quality of life wins for #rstats!

tidyverse.org/blog/2026/02...

Edgar all that work, including heroic efforts to translate TI calculator code to SQL for *prediction intervals*.

I didn’t believe it was possible until he did it. 🍌🍌🍌

caretForecast Conformal Time Series Forecasting Using Machine Learning

The hexagon here is priceless 😎

taf-society.github.io/caretForecast/

#rstats #timeseries

Tomorrow at the Data Science Lab 🧪 we are hearing from the amazing @theotheredgar.bsky.social about the {mall} package:

Run Natural Language Processing against your #RStats tibbles or #Python Polars DataFrames for sentiment analysis, text summaries, and more!

Join us at 12 pm ET: pos.it/dslab