lather, rinse, repeat
on reproducibility
On Thursday, The Washington Post put out the first analysis of its own demographics in the organization’s 143-year history. While the numbers are in many ways disheartening, the transparency is a good step in potentially changing things for the better across the whole organization.
But while the results of the analysis are now public, the analysis itself is not. There are some uncertainties about data classification and methodology that cannot be answered by reading the study itself.
Last year, The Post Guild set out to conduct a study of its own into pay equity at the organization. Over the years, the Guild has studied pay in the newsroom repeatedly but much of what I could find on those previous studies was just the final reports detailing what had been found.
But as the person taking the lead on doing the analysis portion of the study last year, I wanted to know how they’d come to those conclusions and what hit the cutting room floor. I gleaned a little of this from talking to a few folks who had previously studied pay at The Post. But ultimately I had to start from scratch.
Throughout the months of working through it, I wrote all my code in a Jupyter Notebook. From time to time I’d rearrange the code into logical sections so I could find what I wanted if I needed to go back to edit things or look for results. But keeping my code in a notebook had other uses. For one, it allowed others to look at my code and evaluate analysis on their own. But more importantly for me was allowing the Guild to reproduce the analysis in future years quickly. If you know you’re going to repeat something, it’s worth taking extra time to build something that allows future iterations to match the first.
I also rewrote all of the code in R and posted it all to Github so that our methodology could be scrutinized. We also put it all there so that the code could be repurposed for other news organizations who want to do a study of their own without reinventing the wheel. I figured if I had to go through all this work it didn’t make sense for others to start from scratch.
I pride myself on building one-of-a-kind pieces for clients of my woodshop. It’s important to me that folks who want furniture or wood pieces get something they cannot find exactly in someone else’s home.
But sometimes I wind up with a project that has me build a bunch of similar things that require precision to come out similar enough to one another. This was the case when I made four iterations of the same desk in June for a pharmacy in Virginia. There was a lot of repetitive work but most of it was just time-consuming and not too difficult to repeat.
But a current project that requires repetition has been vexing me. The project is to build a bunch of plant cradles inspired by Hilton Carter (as seen below) for my colleague Jenn Abelson. I’m nearly finished after months of on-and-off attempts to get it right.
Part of why it took so long is I didn’t think about making a jig for drilling the holes until recently. In woodworking, a jig is a mechanism that you build for a specific task you need to repeat to provide accuracy and interchangeability when switching in a new item. I don’t have much of a use for hyper-specific jigs because I don’t often repeat, so I didn’t really think about it. If you know you’re going to repeat something, it’s worth taking extra time to build something that allows future iterations to match the first.
So I built a jig for my drill press that holds each cradle in place and allows me to slide pieces to specific spots in the jig so I can drill new holes in the correct places. It also holds the wood tightly so the forstner bit doesn’t send the wood flying while it drills through the uneven live-edge portion of the wood. I ruined a few pieces because that part of the jig wasn’t right, so I went back and tweaked until I got it right.
And now with it right, I can repeat the process to drill all 128 holes with speed and accuracy to build 16 cradles with evenly spaced holes across all of them.
datawork
The NYPD Files — “After New York state repealed a law that kept police disciplinary records secret, ProPublica sought records from the civilian board that investigates complaints by the public about New York City police officers. The board provided us with the closed cases of every active-duty police officer who had at least one substantiated allegation against them. The records span decades, from September 1985 to January 2020. We have created a database of complaints that can be searched by name or browsed by precinct or nature of the allegations.” [ProPublica]
Vallejo police bend badges to mark fatal shootings — “Of the 51 current and former Vallejo police officers who have been involved in fatal shootings since 2000, at least 14 had their badges bent by a colleague afterward, sources familiar with the tradition confirmed. One source told Open Vallejo the number could be much higher. Steve Darden, Vallejo’s most lethal officer of the past 20 years, said allegations he bent his badge are “a lie.” He is seen here in a Mar. 24, 2020 Vallejo Police Department Facebook post announcing his promotion to lieutenant. Vallejo’s shooters typically bend their badges at the 3 o’clock and 4 o’clock points.” [Open Vallejo]
The NYPD Is Still Overwhelmingly Ticketing People Of Color For Drinking In Public — “But while New Yorkers of all stripes are enjoying wine, beer, and cocktails in parks and on piers and stoops, public data shows NYPD officers are disproportionately ticketing Black people and other people of color for drinking in public. Of the 1,250 criminal summonses for drinking in public that the NYPD has handed out since January, 48 percent were given to Black people, and about 43 percent went to Hispanic people. Only 7 percent were given to white people.” [WNYC]
A ‘blight’ of domestic violence deaths strikes Alaska villages — “ The combined population of the villages where the killings occurred is fewer than 1,800 people. In Alaska’s largest city of Anchorage, home to a population 160 times larger, there have been just six homicides in 2020, as of last week. Statewide, Alaska has the nation’s highest rates of sexual assault and women killed by men. The Alaska Justice Information Center published a study in May that found more than 29% of all homicide victims in the state are Alaska Native, yet Alaska Natives make up just 16% of the population.” [Anchorage Daily News]
Pandemic’s weight falls on Hispanics and Native Americans, as deaths pass 150,000 — “And now, the novel coronavirus has officially killed more than 150,000 people in the United States, according to data gathered by The Washington Post. While the disease continues to kill the oldest with impunity, other disturbing trends have emerged. Among them: Hispanics make up an increasing proportion of covid-19 deaths. More than 25,800 have been struck down by the merciless pathogen, which now accounts for 1 out of every 5 deaths among Hispanics, according to data from the Centers for Disease Control and Prevention analyzed by The Post.” [The Washington Post]
woodwork
