View this email in your browser
2019-03-20 | J++ Nyhetsbrev #71

Election compass transparency

In case you missed it: When Helsingin Sanomat, Finland's leading newspaper, received questions about their election compass, they decided to open up the source code.
This allowed people to point out flaws in their model, and give suggestions for how to improve it. After less than a week the algorithm was updated, and hugely improved
At the same time, they also published the raw data in an Excel file.
We predict that this will be used as an example of the advantages of opening up newsroom algorithms in conferences and journalism schools for a long time to come. (Finnish)

As for the flaws pointed out in the HS elections compass, it basically boiled down to small parties being to favourably treated, because their politicians are more homogeneous in their replies.


We did not attend the NICAR data journalism conference this year, but the digital specialists from the Swedish public service broadcaster Sveriges Radio (Digitala insatsstyrkan) did, and here is their excellent write-up:

Inspired by the sport coverage of FiveThirtyEight, J++ Jens Finnäs dug deep into biathlon data to understand who the favorites would be in the world championships. The result was a statistical model for predicting biathlon results. How did the model do?
We gave the Hanna Öberg, the Swedish top runner, a 3–5 percent chance of victory in each run. Not much, but she ended up beating the odds with a gold medal. Here is Jens take on what made that win possible:
Story in Aftonbladet: (Swedish)
Source code:

The next one is from last year, but because we are welcoming Spanish Civio as a new member to the European Data Journalism Network (EDJNet), we'd like to highlight their cross-examination of contraceptives access and usage across the globe.


The Mozilla secure file sharing service is now publicly available (it has been in beta for quite some time). It's a great way to receive files from people who are not comfortable with or in a position to encrypt files themselves or send them though e.g. Signal.

If you are interested in diving into Youtube data (meta data and contents alike), definitely have a look at the Github repository De Correspondent has set up as part of an ongoing research. It's all Python, and there are a lot of scripts there to draw inspiration from.

Want to share a chart as an animated gif? Google's Data Gif Maker can help you there, but be advised that virtually no styling or even basic number formatting is available (yet, at least).
Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.