| | Power BI and Excel are still where a lot of analysis happens, even in Snowflake-based stacks. This webinar shows how teams keep those tools fast and consistent with a shared semantic layer, so analysts can explore large datasets without rebuilding metrics or extracts every time.
*Message from this week's sponsor, AtScale |
| | |
| | If you’re using AI in your day-to-day data work, I’m collecting examples of what’s actually working. Hit reply with a few lines on your workflow, tools, and any surprises or failures. I’ll turn the patterns into a roundup of real-world workflows for Data Elixir readers. |
| | |
| |
|
| A/B testing in Python sounds simple until you try to standardize results across experiments. This post compares SciPy, statsmodels, Pingouin, and tea-tasting through a practical lens and shows how much work it actually takes to get clean, repeatable outputs.
Evgeny Ivanov |
| | |
|
| Outliers aren’t always in the tails. This paper proposes using surprisals to detect anomalies anywhere in a distribution. By converting complex data into a univariate surprisal score, the method makes anomaly detection more robust and easier to analyze.
Rob J Hyndman, David T Frazier |
| | |
|
|
| There's a nice tradition of using sports data to teach serious stats. This one fits squarely in that lineage. It’s a hierarchical Bayesian model built on real disc golf putting data, with full Stan code and proper predictive checking.
Aki Vehtari |
| | |
|
| P-values feel a lot less mysterious when you can build them yourself. By simulating null worlds, this interactive guide makes hypothesis testing something you can see and reason about, not just compute.
Andrew Heiss |
| | |
|
| This post walks through the complete materials from a hands-on workshop on agent-assisted data workflows, covering querying, cleaning, scraping, and visualization with Claude Code and Codex. The focus is on end-to-end interaction patterns, grounded in Python and SQLite.
Simon Willison |
| | |
|
|
| Kevin Markham has taught thousands of practitioners how to use scikit-learn through Data School. His new book distills that experience into a practical guide to the ML workflow, covering data prep, pipelines, model evaluation, and the habits that lead to reliable results. Free to read online.
Data School | Kevin Markham |
| | |
|
|
| Skore aims to make good ML practices easier to follow. Feed it a scikit-learn model and dataset and it produces evaluation reports, visual diagnostics, and cross-validation summaries automatically, turning routine model analysis into a one-line step.
:probabl. | François Méro |
| | |
|
| AI coding assistants can make you productive fast, but they can also make you forget how things work. This Claude Code skill inserts small learning exercises into your workflow, prompting you to predict, explain, or implement pieces yourself so AI-assisted coding still builds real expertise.
GitHub | Dr Cat Hicks |
| | |
|
|
| Before ggplot and Altair, there’s paper, glue, and imagination. This global contest highlights how intuition for data starts with creativity, not code.
Data2Kids |
| | |
|
|
|
This week's issue sponsored by: |
|
|
