"Irish Times Data" - Rigorous?

xrp

Front page of today's "Paper of Record":

irishtimes.com/news/ireland/irish-news/road-deaths-highest-in-counties-with-fewest-penalty-points-1.2151032

Am I right in saying that this kind of stuff is like fingernails on a blackboard to statisticians?

And where is this "Irish Times Data"? Is this data open for inspection?

Gone West

xrp wrote: »

fingernails on a blackboard

Spot on. Not only is their data collection embarrassingly bad, but they clearly try to push an agenda with every weak "conclusion" they pose.
They also don't understand the concept of causation.

xrp

Fuzzy wrote: »

Spot on. Not only is their data collection embarrassingly bad, but they clearly try to push an agenda with every weak "conclusion" they pose.
They also don't understand the concept of causation.

I see they tacked this bit to the end of the article:

"* Note: We found a negative correlation of - 0.419, significant at 0.05 level, between the rate of penalty points in each county and road deaths per county per 100,000 population."

Reassuring?










*** Edit: the "negative correlation of -0.419" bit appears to have been memory holed...

Black Swan

xrp wrote: »

I see they tacked this bit to the end of the article:

"* Note: We found a negative correlation of - 0.419, significant at 0.05 level, between the rate of penalty points in each county and road deaths per county per 100,000 population."

Reassuring?

Well, I have not read the article, and am just shooting from the hip here, but if I am observing what you offered here correct, it would appear that there could be a directional negative hypothesis: (HR) As rate of penalty points increase, road deaths per county decrease. The null to be tested would be no relationship between variables.

The results of p<.05 would be significant for most disciplines, but the -0.419 would render an explained variance of only 17.6, leaving 82.4 unexplained, suggesting that variables not measured (or not mentioned here) in this research accounted for most of the variation. Given that correlation is a necessary but insufficient condition to establish causation, I would want to review the conceptual framework, methods, data, as well as other sources of information in an attempt to rule out spuriousness, etc.; i.e., the information here is insufficient to suggest a tentative conclusion.

Then again, I am posting this at 5AM Pacific, have pulled an all nighter, am wired but tired, and may be talking complete nonsense accordingly.

xrp

So I did some digging of my own...

Between 09:00 and 10:00 this morning:

diff-online.com/view/55117c5c6072b

Between 10:00 and 12:00 today:

diff-online.com/view/55117c26d252a

Between 12:00 and 13:00 today:

diff-online.com/view/55117c8f17fe7

These are not just minor changes (spelling, grammar, times, etc.) that is common in "breaking news" or "developing stories" type articles. It's a blatant back-track by the editors. I wonder will the Irish Times publish a clarification? Or are they just hoping nobody will notice?

The print article that appeared on the front page around midnight last night (cf.broadsheet.ie/wp-content/uploads/2015/03/it15.jpg) is unrecognisable from what's currently online. The only thing that's the same is the headline.

And people are supposed to pay a premium for this?

Ouch.

Gone West

Nice catch. I'm not a statistician but from what I understand, these guys are selling articles for clicks and don't even pretend to pay attention to detail any more.
As you can see from the article, a "data scientist" and a social affairs "journalist" share the byline. It seems that they need to teach their data people to write, as teaching the journos science is clearly not working.