ISSUE 351 · August 31, 2021
Trends
The Modern Data Experience
A lot of "normal" people work with data these days and most don't care about the stack of tools that data practitioners use to move, store, wrangle, analyze and present data. They care about their own experience with data. "...the modern data stack isn't enough. We have to create a modern data experience." Great post!
Benn Stancil
Sponsored Link
Using third-party data to make smarter decisions
The insights gained by using third-party data with your internal data can help you make smarter business and technical decisions. This technical eBook can help you learn how to use AWS Data Exchange to couple data sets with advanced analytics and machine learning with step-by-step instructions.
Tutorials, Projects & Opinions
A lightweight data validation ecosystem
Nice approach for building an "advanced" but right-size data monitoring solution using common tools: GitHub (Actions, Pages, issues), R (pointblank + projmgr pkgs), and Slack notifications.
Emily Riederer
Inferring Concept Drift Without Labeled Data
A trained model might be static but because people and systems are constantly changing, models tend to drift over time. Detecting drift is key but that can be problematic with unlabeled data. This research report explores the problem and offers four ways to approach it using unsupervised methods.
Cloudera Fast Forward Labs
Pseudo-R²: A Metric for Quantifying Interestingness
Great post that shows how the data science team at Heap uses a statistical metric called pseudo-R² to quantify how interesting an insight will be before deciding to recommend it for analysis.
Heap | David Robinson
Science For People Who Give A Shit
Join tens of thousands of other smart humans and subscribe to Important, Not Important for the most vital science news of the week, deep analysis, and Action Steps you can take to 1) feel better and 2) measurably improve the world around you. Get it for free.
// sponsored
Code & Tools
SQLModel
SQLModel is a library that makes it easy to interact with SQL databases from Python code. It's designed to be intuitive, highly compatible, and robust. SQLModel is based on Python type annotations and is built as a layer on top of Pydantic and SQLAlchemy.
SQLModel | Sebastián Ramírez
Resources
Analyzing US Census Data: Methods, Maps & Models
Kyle Walker is a spatial data science expert who specializes in population geography and demographic trends. In this new book, Kyle shows how to acquire, wrangle, visualize, and model Census data using R. This looks like a great resource and is free to read online.
Kyle Walker
Data Visualization
prettymaps
pretty maps is a small Python library to draw pretty maps from OpenStreetMap data. Based on osmnx, matplotlib and shapely libraries.
GitHub | Marcelo Prates
Data Vis Dispatch
Datawrapper's new data visualization newsletter is a great source for curated visualizations from the media. Along with showing how data is being used to portray current events, many of the picks are good examples of visualization techniques. Sign up at the bottom of the page.
Datawrapper | Rose Mintzer-Sweeney and Lisa Charlotte Rost
 |