Stats in the newspapers
As I’m waiting to catch my flight back from Stockholm (where I had a rather interesting meeting discussing Bayesian network meta-analysis. But I digress…), I am browsing the news — mainly just to keep myself awake. And I find this interesting piece of news in The Guardian. It is the story that data released by the Nigerian Electoral Commission (INEC) and then analysed by the journalists show that
for each of the country’s 36 states and its capital shows that INEC has increased the number of new registered voters by almost exactly the same percentage across all states. The correlation is a “statistical impossibility” and does not reflect Nigeria’s demographic changes, according to data analysts working with the Guardian. Additional data seen by the Guardian also shows irregularities in registration for the 2015 election, until now considered to have been free and fair.
and that
Plotted on a scatter line graph, there is a 0.99 correlation across all the states, without a single outlier. According to three separate data analysts, the parity cannot be a coincidence. “Only God works that closely,” one analyst said.
Now, I think this is interesting — but it would have been even more interesting if the article actually showed the graph. And for that matter, if the data were made available publicly (I’m assuming the information would be presented in an aggregated form anyway, so should be possible to disseminate without particular bridge of privacy, etc?). It may be my state of slightly sleep-deprivation (XX has not been well, lately, which means she’s not slept well, which means none of us has…), but I don’t think any link was provided — which I also found a bit odd, as places like The Guardian usually do so.
And while I’m here, it just occurred to me that last week I saw another interesting news-related story. This is about how the BBC Visual and Data Journalism team have modified their workstream and now analyse their data and produce the graphs that eventually get published on the BBC news website using R and ggplot2. Which I think is pretty cool. In fact, they even went as far as to create their own package (called bbplot), in which they have customised the standard ggplot2 themes to adhere with their own formats. Which, again, I think it’s pretty cool!