Some Pitters and Patters with R | Simon P. Couch – Simon P. Couch

How I’m using Claude Code to write R code

Giving Claude Code the ability to peruse R package documentation via MCP is so, so helpful.

Kimi K2 and R Coding

An open-weights model released over the weekend by a little-known company has drawn quite a bit of attention. Is it any good?

Claude 4 and R Coding

Evaluating the most recent releases of Claude Sonnet and Opus on challenging R coding problems.

Evaluating Gemini 2.5 Flash on R coding tasks

Google’s Gemini 2.5 Pro release really made a splash last month. They’ve just announced an update to 2.5 Flash, a faster and cheaper model.

Evaluating the new Gemini 2.5 Pro update on R coding

The initial Gemini 2.5 Pro release a month ago was surprisingly strong, so I was excited to benchmark the update announced yesterday on R coding problems.

Evaluating o3 and o4-mini on R coding performance

With a pair of coding-focused model drops, it’s clear OpenAI is aiming for the developer market. How well do the newest o3 and o4-mini perform on R coding tasks?

How good are the GPT 4.1 models at writing R code?

OpenAI dropped a trio of models yesterday that they claim result from “focusing closely on real-world developer needs.” I was curious whether this might mean stronger R coding performance.

Introducing chores

A package to help with repetitive but hard-to-automate data science tasks.

How Good Is Gemini 2.5 Pro at Writing R Code?

Since Gemini 2.5 Pro’s release last week, I’ve been seeing a lot of hype claiming that the model is the new state of the art. How well does it know R?

How I’m Using Claude Code to Develop R Packages

I’ve come to really appreciate having Claude Code as part of my LLM toolkit, but it’s taken a bit of getting used to.

Exploring Biases in GPT-4o, Claude, and Qwen2.5 Judgements

What happens when you ask language models to evaluate their own output?

Which names that are also names of countries are most common?

Sitting around the living room over the weekend, someone asked this question. Let’s poke at an answer.

Announcing a new book

Today, during R/Pharma 2024, I open-sourced the draft of a book I’ve been working on.

Postprocessing is coming to tidymodels

The tidymodels team has been hard at work on postprocessing, a set of features to adjust model predictions. The functionality includes a new package as well as changes across the framework.

A new package for profiling parallel R code

Periodic snapshots of CPU and memory usage consumed by R sessions enable better analyses of parallel computation.

How to best parallelize boosted tree model fits with tidymodels

Should tidymodels users use the parallelism implementations from XGBoost and LightGBM?

Measuring elapsed time to fit with tidymodels

The development versions of tidymodels packages now include tools to benchmark training time.

Run an Oracle Database with Docker Desktop on ARM (M1, M2, M3) MacOS

Starting a containerized Oracle database on an Apple Silicon Mac.

Analyzing my own music listening data with R and the tidyverse (2023)

All of my friends are posting their Spotify Wrapped, and I’m jealous.

Predicting flight delays with tidymodels🛩

Trying out a variety of machine learning models to predict flight delays out of Madison, WI.

Quarto!

I recently rebuilt this site with Quarto.

Down the submodels rabbit hole with tidymodels

Debugging something that turned out not to be a bug at all.

Hiking the 2023 John Muir Trail / Nüümü Poyo

A group of buddies and I walked 242 miles from Mt. Whitney to Yosemite Valley.

Optimizing model parameters faster with tidymodels

A couple small changes can greatly speed up the hyperparameter tuning process with tidymodels.

Moving On From Baltimore

After 2 years in Baltimore, I recently made the move to Chicago, IL.

The tidymodels is getting a whole lot faster

Recent optimizations have made fits on small datasets much, much snappier.

Analyzing my own music listening data with R and the tidyverse

All of my friends are posting their Spotify Wrapped, and I’m jealous.

Redirecting from sub-domains with Netlify

After merging two websites that deployed from different branches, I had to figure out how to redirect my old blog posts to their new home.

{infer} v1.0.0 is on CRAN

Some short reflections on working on the {infer} R package.

Applying to the NSF Graduate Research Fellowship (GRFP)

A compilation of the advice that made me feel most informed and confident while applying for the NSF Graduate Research Fellowship.

Pipe-esque Programming with {ggplot2}’s Plus Operator

Writing iterative code with ‘+’ rather than ‘%>%’ was a tough transition my first time around.

Big Things (Developer Documentation pt. 4)

On the tension between documenting R packages exhaustively and maintainably.

Naming the Things (Developer Documentation pt. 3)

Weighing the pros and cons of several possible schemas for naming the core functions in {stacks}.

Splitting Things Up (Developer Documentation pt. 2)

Why {stacks} requires (at least) four separate functions to build an ensemble model rather than wrapping them all up into one.

How {stacks} Came To Be (Developer Documentation pt. 1)

Introducing a set of blog posts reflecting on the development process of the {stacks} package.

What’s Next For Me

Applying to Graduate School in Statistics and Biostatistics

There were all sorts of tips and tricks about applying to grad school that I was only able to learn about via office hours, personal meetings, and Twitter DMs, and I thought it would be worth publicly compiling some lessons learned.

Running R Scripts on a Schedule with GitHub Actions

Some pointers on running R scripts and committing their results to a GitHub repository on a regular interval using Actions.

{stacks} v0.1.0 is on CRAN!

Introducing ensemble learning to the tidymodels.

Programatically Generating PDF Reports with the Tidyverse

In a few different roles over the past few years, I’ve come across the problem of programatically generating some kind of PDF reports from data. Here are some tips/tricks I’ve come across while making that happen.

A Gentle Introduction to Tidy Model Stacking

Model stacking is an ensembling technique that involves training a model to combine the outputs of many diverse statistical models. The stacks package implements a grammar for tidymodels-aligned model stacking.

The Blog Thing! We’re Doing It!

For a while now, I’ve wished I had a place to share some short posts on what I’m up to related to R, data science, and statistics. I decided I’d take a Sunday to get this subdomain up and running!

Reuse