Data Journalists Discuss Their Tools of Choice
Psssss. That is the sound of your data decompressing from its airtight wrapper. Now what? What do you look for? And what tools do you use to get stuck in? We asked data journalists to tell us a bit about how they work with data. Here is what they said.
Lisa Evans, The Guardian
At the Guardian Datablog we really like to interact with our readers and allowing them to replicate our data journalism quickly means they can build on the work we do and sometimes spot things we haven’t. So the more intuitive the data tools the better. We try to pick tools that anyone could get the hang of without learning a programming language or having special training and without a hefty fee attached.
Cynthia O’Murchu, Financial Times
Am I ever going to be a coder? Very unlikely! I certainly don’t think that all reporters need to know how to code. But I do think it is very valuable for them to have a more general awareness of what is possible and know how to talk to coders.
Scott Klein, ProPublica
The ability to write and deploy complex software as quickly as a reporter can write a story is a pretty new thing. It used to take a lot longer. Things changed thanks to the development of two free/open source rapid development frameworks: Django and Ruby on Rails, both of which were first released in the mid-2000s.
Cheryl Phillips, Seattle Times
Sometimes the best tool can be the simplest tool — the power of a spreadsheet is easy to underestimate. But using a spreadsheet back when everything was in DOS enabled me to understand a complex formula for the partnership agreement for the owners of The Texas Rangers — back when George W. Bush was one of the key owners. A spreadsheet can help me flag outliers or mistakes in calculations. I can write clean-up scripts and more. It is a basic in the toolbox for a data journalist. That said, my favourite tools have even more power — SPSS for statistical analysis and mapping programs that enable me to see patterns geographically.
Gregor Aisch, Open Knowledge Foundation
I’m a big fan of Python. Python is a wonderful open source programming language which is easy to read and write (e.g. you don’t have to type a semi-colon after each line). More importantly, Python has a tremendous user base and therefore has plugins (called packages) for literally everything you need.
Steve Doig, Walter Cronkite School of Journalism of Arizona State University
My go-to tool is Excel, which can handle the majority of CAR problems and has the advantages of being easy to learn and available to most reporters. When I need to merge tables, I typically use Access, but then export the merged table back into Excel for further work. I use ESRI’s ArcMap for geographic analyzes; it’s powerful and is used by the agencies that gather geocoded data. TextWrangler is great for examining text data with quirky layouts and delimiters, and can do sophisticated search-and-replace with regular expressions. When statistical techniques like linear regression are needed, I use SPSS; it has a friendly point-and-click menu. For really heavy lifting, like working with datasets that have millions of records that may need serious filtering and programmed variable transformations, I use SAS software.
Brian Boyer, Chicago Tribune
Our tools of choice include Python and Django. For hacking, scraping and playing with data, and PostGIS, QGIS and the MapBox toolkit for building crazy web maps. R and NumPy + MatPlotLib are currently battling for supremacy as our kit of choice for exploratory data analysis, though our favorite data tool of late is homegrown: CSVKit. More or less everything we do is deployed in the cloud.
Angélica Peralta Ramos, La Nacion (Argentina)
At La Nacion we use:
Pedro Markun, Transparência Hacker
As a grassroots community without any technical bias we at Transparency Hackers use a lot of different tools and programming languages. Every member has it’s own set of preferences and this great variety is both our strength and our weakness. Some of us are actually building a ‘Transparency Hacker Linux Distribution’ which we could live-boot anywhere and start hacking data. This toolkit has some interesting tools and libraries for handling data like Refine, RStudio and OpenOffice Calc (usually an overlooked tool by savvy people, but really useful for quick/small stuff). Also we’ve been using Scraperwiki quite a lot to quickly prototype and save data results online.