Recent Writing

TSA Screening Volume and Epiweeks

I drew a picture of Subway ridership volume yesterday, in the wake of some absurdly disingenuous official statements about New York City’s congestion pricing scheme somehow restricting New York to a “small elite”. The Subway has higher daily throughput than every TSA Airport in the United States combined. (And, even now, ridership remains down from pre-COVID levels.) So I put in TSA traveler screening volumes on the MTA graph for comparison, just to help make the point.

MTA Ridership

Burn Notice

Your Phone and Watch have a lot of data about you. I mean, like, a lot. Someone should really write a book all about the general issues for society that this raises. Yesterday I decided I wanted to take a look specifically at the health data on my iPhone. I’m not a huge user of the iPhone’s or the Apple Watch’s health features. I don’t use or subscribe to Apple Fitness+, for example. I’m not in any studies. But I do have a bathroom scale that records data and I allow the Watch to keep an eye on my activity. This means that, like so many people, I have grown to heartily detest the blandly affirming Californian inside those devices who periodically encourages me to take a walk, or stand up, or be mindful, and so forth.

Kerning and Kerning in a Widening Gyre

This post summarizes an extended period of deep annoyance. I have tried to solve the problem it describes more than once before and not quite done it. This has, in fact, happened again. I have still not satisfactorily solved the problem. But this time I know why I can’t solve it in a civilized manner. My goal is simple, and reasonable. I want to produce more or less identical plots in both PNG and PDF formats. PNG is a raster format. PDF is a vector format and also the Devil Incarnate. Sometimes you want one format, sometimes the other. Raster formats color in pixels on a grid of some fixed resolution. They are efficient when you need to plot a lot of elements, but you can’t zoom in on them without loss. When you make one, any as it were “structural” information about plot elements is lost. A line or a shape no longer exists as an editable line or shape. It’s just pixels. Vector formats can be easily resized up or down without loss of fidelity and keep more of the structural information used to make the plot to begin with. Lines and shapes remain lines and shapes. But vector formats get big real fast when you have a lot of objects to show, because each one is drawn separately. Also they are the Devil Incarnate. Especially when it comes to one special subset of lines and shapes: fonts.

Halloween Data Cleaning

This week in Modern Plain Text Computing we put together some of the things we’ve been learning about cleaning and tidying data. Here’s a somewhat sobering example using data from the Fatality Analysis Reporting System, which is how the NTSA tracks information about road accidents in the United States. Our data file shows counts of pedestrians aged <16 killed in road accidents on each day of the month in the United States from 2008 to 2022.