31 March 2016

Nursery rhyme conundrum

xkcd picks up a technical issue in one of our canon of nursery rhymes – although it seems it was likely not to have been intended so literally when it was first sung. I have probably sung this hundreds of times both as a child and a parent but never thought beyond the words to the improbability of the scenario.

25 July 2014

A different view of 'The Facebook study'

Michelle Meyer writes in defence of the Facebook/Cornell study on people's response to the emotional tenor of their Facebook feeds. She argues the manipulation of individuals' news feeds wasn't much beyond the everyday experience of Facebook and, yes, a little ethical review and participant briefing might have been nice but let's not let over-sensitivity get in the way of corporations doing what, apparently, they must.

via Ed Yong

29 June 2014

London Underground – re-signed

This set of images from Prosign has, apparently, been around for some time but I've only just seen it.

via Alex Pang on Facebook

24 June 2014

The sun always shines on...

... the Department of Typography & Graphic Communication, University of Reading (Class of 2014).

06 June 2014

Gangnam opportunity cost

The Economist has mapped the aggregated time spent watching 'Gangnam Style' on YouTube to other projects which have demanded much human time and effort.

Via Tim Harford

22 April 2014

10,000 hours of practice – unpicked

Nice pop explanation at Salon of the need for quality, spaced practice, not just massed practice. This is one of my favourite, non-intuitive and robust psychological phenomena, first demonstrated in the late 19th century (by Ebbinghaus) and replicated in multiple studies subsequently.

[via Katja Battarbee]

16 April 2014

Of marginal significance...

Have been wondering about the term 'marginally significant' used in research papers. It's a term that wasn't used when I learned statistics a very long time ago. Research results were either significant, at a given level, or not. So I browsed the term and found it is, indeed, acceptable in some settings, for reporting 'non-critical results' (you define what's critical).

Then I found this lovely analysis by epidemiologist, Matthew Hankins, of the p values at which effects are described as marginally significant. I love the annotations.

 And some detailed discussion by Hankins here.

Also a very nice video, 'Dance of the p Values', by Geoff Cumming which starts in the same vein has Hankins' diagram.