Below you’ll find a recent example of the monthly newsletter. Click here to subscribe.


Analyse Garmin or Strava activities using Python

Last year, I analysed how crosswinds affect cycling speed. As a spinoff, I created the bikeride Python package to analyse gps-recorded activities. I’ve now written an article on how to use this package.


Student files complaint against racist surveillance software

Universities use software like Proctorio to monitor students who take online exams at home. This type of software, likened to spyware, uses technologies like facial recognition and eye movement detection to identify students and detect ‘suspicious’ behaviour.

Robin Pocornie, a student at the Vrije Universiteit Amsterdam, found that she had to shine a light into her face for the software to recognise her as a human being, a problem her white fellow students didn’t have. It’s well-documented that facial recognition software may not perform well with non-white faces, but this hasn’t stopped universities from using this technology. Pocornie has now filed a complaint with official human rights watchdog College voor de Rechten van de Mens (via).

Tesla sensors

On a somewhat lighter note - newspaper NRC Handelsblad had a rather hilarious article (paywall) about the anarchist Red Bicycle collective. They target VanMoof bicycles - which they consider a symbol of late capitalism - but are also considering other targets. «‘Did you know that each Tesla has eight cameras and twelve sensors attached to it which are always on? If that doesn’t constitute illegitimate appropriation of public space... But how to tackle this?’ ‘Perhaps with mirrors,’ Mike says. ‘So the cameras capture themselves. Like that James Bond movie, with the invisible car’.»


Attempt to file an FOI request - in person

Dutch voters want the minimum wage raised to at least 14 euro and have been in favour of measures to reduce income ineqality for decades. Yet government decisions fail to reflect what voters want. Trade union FNV decided to find out what drives government decision making, by filing a freedom of information request. Not just by sending an an email - activists tried to deliver the request in person. The office of the Prime Minister declined to accept it, and sent the police.


Colour scales for weather maps

As temperatures rise, MapLab discusses the colour scales used to show temperatures on weather maps. These scales should be inituitive, but also avoid greens, which many colour-blind people can’t perceive. The British Met Office has redesigned its scale, eliminating greens and adding brightness, with darker colours for extreme temperatures. Arguably, this also resulted in prettier maps (the entire thread is worth a read).

This discussion might be of interest to Dutch weather institute KNMI as well.

Amsterdam maps 220 thousand historical addresses

Amsterdam has created a website (article, search) where you can look up addresses found in historical sources. The addresses are linked to their location, and sometimes to other addresses that have been found for the same location. You can also download the entire dataset.

Interestingly, for some time, addresses weren’t identified by their street name. Instead a combination of a neighbourhood code, consisting of one or two letters, and a house number was used. This article tells the story of the less than smooth introduction of addresses in Amsterdam.

Tour stage visualised

Nice combination of an elevation profile and a map, visualising three Tour stages held on 27 July 1939. Unfortunately, the source is not named.

Misleading graphs

How misleading is a truncated y-axis

A group of Dutch academics have analysed the effect of ‘misleading graphs’. Respondents were presented with a text and a line chart about the ‘Bluebeak’, a fictional bird species. In some cases the text said that the Bluebeak is endangered; in other cases the text said that the bird causes ecological disruption. The line chart showed growth of the Bluebeak population over time. In some cases, the y-axis started at zero; in other cases it started at 2,400 (the charts can be found in the article).

Respondents were asked for a judgement about the situation (good or bad), and to estimate the percentage increase of the Bluebeak population.

The graph type didn’t make much difference for respondents’ judgement of the situation, which was largely dependent on the text they were presented with. This makes sense: regardless of whether it is interpreted correctly, a chart can’t tell you much about how good or bad a situation is if you don’t know the context.

The design of the y-axis did make a substantial difference for respondents’ estimates of the population increase. On average, respondents who had seen a ‘normal’ chart thought the population had grown 19% (which is about correct), while people who had seen the chart with the shortened y-axis on average thought it had grown 29%.

The authors found that the influence of graph type on estimates disappears if you take ‘graph literacy’ (that is, someone’s ‘ability to read, process, and comprehend data visualisations’) into account. This led them to conclude that «people are unfazed by misleading charts — except for people with a low graph literacy when it comes to making estimations.»

This conclusion about graph literacy is based on a multiple regression analysis. However, a more basic analysis (using the data the authors published with the article) raises the question whether people with higher graph literacy really are immune to the effect of a shortened y-axis. It appears that for each level of graph literacy, people shown a chart with a shortened y-axis make higher growth estimates than those shown chart with a y-axis starting at zero.

In any case, if you’re using a chart with a shortened y-axis, you might want to give a visual hint so people are made aware of this.