The New York Times provides a breakdown of minutes spoken at the Republican National Convention. The bubbles, sized by minute count, start as an overview of everyone who spoke, and then cluster into specific groups as you scroll.

Bloomberg looks at how retail struggles might kill the middle-of-the-road malls before this pandemic is done:

Although many bankrupt retailers continue operating while restructuring under Chapter 11, they’re planning to shut down droves of lower-performing stores. Justice recently shuttered its location in Crystal Mall after its parent company, Ascena Retail Group Inc., filed for bankruptcy on July 23. The mall also houses a Men’s Wearhouse, whose parent, Tailored Brands Inc., filed for bankruptcy on Aug. 2. It wants to close up to 500 stores, accounting for a third of its locations. Vitamin retailer GNC, which filed for bankruptcy on June 23, wants to close at least 800 to 1,200 stores. They both operate in Crystal Mall.

I like these triangles to show scale. There’s also a variable width bar chart in the piece. It’s so back.

“Two.js is deeply inspired by flat motion graphics. As a result, two.js aims to make the creation and animation of flat shapes easier and more concise.” It also renders in webgl, canvas2d, and svg, with not much change in your code. Two.js is definitely going on my list of things to try.

Here’s the good stuff for August 2020.

Talking about a possible plasma treatment for Covid-19, the Food and Drug Administration Commissioner Stephen Hahn misinterpreted results from the study. The study from the Mayo Clinic notes a possible 35% reduction in mortality rate, and Hahn said that if 100 people were sick with Covid-19, 35 lives would be saved.

For The Washington Post, Aaron Blake discusses why the interpretation is incorrect:

The vast majority of people who get the virus will recover with or without plasma. The 35 percent figure comes into play among those who die — a much smaller group. That would still be a huge development if borne out. But strictly speaking, the treatment would have saved about 3 out of 100 coronavirus patients, not 35. And given the smaller numbers we’re talking about, the finding is much closer to the margin of error — even as the preliminary study finds the effect to be statistically significant.

And even then, the claim doesn’t make sense. The data that he and Trump were referring to compared those receiving plasma treatments not to a control group, but between higher and lower levels of plasma treatments. The group with lower levels died at a rate of 11.9 people out of 100 died, while 8.7 percent died with higher levels.

Hahn later corrected himself.

See also Christopher Ingraham’s quick explanation of relative versus absolute risk. And this visual explainer from 2015 by NYT’s The Upshot should also be helpful in understanding the difference.

How do you assemble a banana and peanut butter sandwich that maximizes the number of bites with the perfect ratio of bread, peanut butter, and banana? Ethan Rosenthal, in a quest to work on something truly meaningless, solved the problem over several months with a truly roundabout solution:

So, how do we make optimal peanut butter and banana sandwiches? It’s really quite simple. You take a picture of your banana and bread, pass the image through a deep learning model to locate said items, do some nonlinear curve fitting to the banana, transform to polar coordinates and “slice” the banana along the fitted curve, turn those slices into elliptical polygons, and feed the polygons and bread “box” into a 2D nesting algorithm.

Best.

For The Washington Post, Sergio Peçanha asks, “What will it take to achieve gender equality in American politics?”

It will take some more time and a lot more effort to reach equal representation. I asked my colleague David Byler, a statistics expert, to estimate how long it would take for women to reach equal numbers in Congress at the current pace. His estimate: about 60 years.

Brad Plumer and Nadja Popovich for The New York Times show how policies that marked black neighborhoods as “hazardous” for real estate investment led to a present-day with fewer trees and higher temperatures. The maps that shift back and forth between past districting and how things are now show the picture clearly.

This goes hand-in-hand with how tree-cover and neighborhood incomes are also tightly coupled.

Businesses are still seeing visits mostly down compared to last year, which shouldn’t be much of a surprise. But there is a lot of variation across the states. The New York Times shows the comparison over time, based on mobile location data (which I still feel uneasy about). NYT went with the scrollytelling state-by-state approach to work their way through the spaghetti plot.

It’s clear that Covid-19 has affected groups differently across the United States. By geography. By education level. By income. The Marshall Project breaks down excess deaths by race:

Earlier data on cases, hospitalizations and deaths revealed the especially heavy toll on Black, Hispanic and Native Americans, a disparity attributed to unequal access to health care and economic opportunities. But the increases in total deaths by race were not reported until now; nor was the disproportionate burden of the disease on Asian Americans.

With this new data, Asian Americans join Blacks and Hispanics among the hardest-hit communities, with deaths in each group up at least 30 percent this year compared with the average over the last five years, the analysis found. Deaths among Native Americans rose more than 20 percent, though that is probably a severe undercount because of a lack of data. Deaths among Whites were up 9 percent.

Difference charts are used to show deaths above (red) or below (turquoise) normal counts, but of course, it’s mostly red.

See the piece for an additional categorization by state.

From the Computer Graphics Lab at Stanford University, the results from an analysis of a decade of cable news:

The Stanford TV News Analyzer has applied deep-learning-based image and audio analysis processing techniques to nearly a decade of 24–7 broadcasts from Fox News, CNN, and MSNBC going back to January 1, 2010. That’s over 270,000 hours of video updated daily. Computer vision is used to detect faces, identify public figures, and estimate characteristics such as gender to examine news coverage patterns. To facilitate topic analysis the transcripts are time-aligned with video content, and compared across dates, times of day and programs.

You can search for topics or people, combine queries, and set time ranges. Then you get a time series for how much someone’s face showed up or the number of times a word was used.

Give it a go.

The volume of mail-in ballots will likely be higher than usual this year, but relative to the Postal Service’s usual volumes from years past, the bump doesn’t seem unfathomable. The chart above, which shows average weekly volume over the years, from Quoctrung Bui and Margot Sanger-Katz for NYT’s The Upshot, shows the scale.

Of course, if certain administrations continue to hamper USPS operations, that’s a different story.

Last month I did a short Q&A about FD and my workflow. I thought I’d elaborate on one of my answers.

Researchers from the National University of Singapore found a way to infer key shape based on the sound the lock makes when you insert the key.

First they capture a sound recording with a standard microphone. Then they run the audio file through software to filter out the metallic clicks. This provides a time series from which they can infer likely keys.

Soundarya Ramesh presented the work at HotMobile 2020 in the talk below:

Oh to be back in graduate school again. [via kottke]

Los Angeles Times provides a California-specific map of the current wildfires to stay updated on what’s happening right now.

In the zoomed out view, hexagons bin the individual fires and color by number of hotspots. Wavy hatching indicates levels of air pollution. In the zoomed in view, see the individual fires and click for current status.

With the rush of wildfires in California, governor Gavin Newsom declared (another) state of emergency. The Fire and Smoke Map from the U.S. Forest Service and Environmental Protection Agency provides a picture of where we’re currently at. The map incorporates data from a variety of sensors across the country:

The sensor data comes from PurpleAir, which crowdsources data from that company’s particle pollution sensors and shows the data on a map. Before the sensor data appear on the AirNow Fire and Smoke Map, EPA and USFS apply both a scientific correction equation to mitigate bias in the sensor data, and the NowCast, the algorithm to show the data in the context of the Air Quality Index.

For The Los Angeles Times, Casey Miller went hyperlocal to track mask wearing in three locations in Los Angeles and Orange counties. Over a week, a group of reporters counted people who passed by and tallied if people wore their mask correctly, incorrectly, or no mask at all.

The above is the breakdown for a spot on Main Street in Huntington Beach.

Maybe the best part is that there’s a simple tool at the end so that you can count in your own spot:

If it weren’t so smoky outside, I’d give this a go.

There was an explosion in Beirut. It was big. How big? Marco Hernandez and Simon Scarr for Reuters provide a sense of scale:

George William Herbert, an adjunct professor at the Middlebury Institute of International Studies Center for Nonproliferation Studies and a missile and effects consultant, used two methods to estimate the yield of the explosion. One used visual evidence of the blast itself along with damage assessments. The other calculation was based on the amount of ammonium nitrate reportedly at the source of the explosion.

Both techniques estimate the yield as a few hundred tons of TNT equivalent, with the overlap being 200 to 300, Herbert told Reuters.

It starts with a Hellfire Missle, which is 0.01 tons. Then it just keeps going.