Zurich R User Group, March 2016

Data journalism with R

Experiences at SRF Data

Timo Grossenbacher

@grssnbchr

@srfdata

Presentation available under grssnbchr.github.io/zrug-rddj

About me

Formerly: MSc in Geography & Computer Science UZH

Mar - Oct 2014: Tages-Anzeiger

Since Nov 2014 I work in the team of SRF Data as coder and journalist ("journocoder")

@grssnbchr

What the heck is data journalism?

Some examples

• NYT: The Best and Worst Places to Grow Up
• ProPublica: Dollars for Docs
• CORRECT!V: Euros für Ärzte

SRF Data

@srfdata

Portfolio

SRF Data

More: research & ideas, less: service center

pitch ideas, receive / collect / scrape / enforce (BGÖ) data

preprocess > visualize > analyze > find the story

publication on srf.ch -> overview, interactivity

publication in radio and/or TV -> anecdotes, details, aspects

Some examples of our work

Dual-use goods & conventional arms exports

The Zurich cantonal police orders a trojan horse

Covering of the Federal Elections 2015

What about R?

R is a Swiss Army Knife > it reads all kinds of weird s**t

R allows for automation > example

R empowers reproducibility and, ultimately, transparency > we publish most of our analyses on GitHub

I wrote a blog post about all of this

How exactly do we use R?

base, dplyr, tidyr, maggritr, ggplot2, extrafont, animation, readxl, xml2, jsonlite, RSQLite, googlesheets, stringr, rpremraj/mailR, R2HTML, knitr, slidify, readr, caret, sp, maptools, etc.

Things I still don't like about R

Error reporting in RStudio - in general, the console in RStudio

JSON

The plethora of packages doesn't make it better (jsonlite, rjson, RJSONIO). You still need to use list() and complicated lapply() calls to produce nested data structures - and JSON is all about nested data structures.

What about something like that?

my_dataset %>% group_by(facet) %>% to_json("output.json")

The language... especially Standard Evaluation vs. Non-Standard Evaluation ... and stuff like paste() or paste0().

direct_matches %<>% mutate_(.dots = 
    setNames(
        list(
            interp(~ as.numeric(sub("\\D*(\\d+).*", 
            "\\1", a)), 
            a = as.name(combined))), 
        combined
        )
    )

Some resources

rddj.info - a resource collection for doing DDJ with R

grssnbchr/rddj-reproducibility-workflow

Brian Keegan calls out 538 for openness in #ddj

The future of (data) journalism?

"Real" interactivity

Sonification

Snackable graphics

Bots and IoT

Algorithmic Accountibility

Thank you!

Questions?

@grssnbchr or timo.grossenbacher(at)srf.ch

@srfdata

This presentation is available (and reproducible) under github.com/grssnbchr/zrug-rddj

Built with the R package slidify and revealjs