statistics and applied probability

national university of singapore

Month: July 2010

The Canadian census

In case you missed it on allstat: the Tory government in Canada has decided to change their census, so that instead of requiring 20% of the population to fill out the “long form”, they will now merely ask (nicely?) for the form to be filled.  And to make up for some people not doing it, they’ll send out more of them, to 30% of the population. So they pay 50% more for something that’s worth much much less, cos they won’t have a clue what causes non-response, and nor will any researchers across their country for the next five years. (And any Canadian interested in historical changes over time for the centuries ahead are going to find a black hole around 2010.)

John Campey and the Data Hounds (including Sarah Palin?) had this to say:

If you, like me, think this decision is totally thick, then you could sign this petition.

Oh, and Nature ran an editorial on this débâcle, which is where I got the link to the petition (and the youtube video).

P.S. It’s not really a census if it’s “just” 20% of the population: it’s a sample. Similarly our own “census” is not a census (I did Singapore’s “census” this year, and got very irate in the process. Not just the name, but the way the answers were designed, it was really really bad. For instance, I had to choose my “race”. Searching through their list, I think I had the choice of {Caucasian, Briton, Scot}, any of which is right, but I only got to choose one, despite Scot being a subset of Briton being a subset of Caucasian.  So they probably have the right number of Scots living in Singapore but one Briton and one Caucasian too few.  Ditto when it came to my wife’s religion, they only let me choose one, but as far as I’ve been able to tell, she has two religions.)

Oracle octopuses

Paul the psychic octopus, taken by Tilla

Paul the psychic octopus, taken by Tilla

Well, Maradona’s dream about Chino lifting the world cup didn’t turn out too well. But if you missed it, Paul the psychic octopus proved to be a better oracle. He predicted the winner of eight games—seven involving Germany as well as the final—and got all eight correct. Simple calculations give a p-value of 0.4% for the null hypothesis that he guessed with probability 50% each time. Hey, maybe he really is psychic?!

Or maybe there are lots of wannabe oracle animals out there and Paul’s the only one we know about.

Or maybe he liked the flag of a (fairly) good team (Germany). Wikip has a variety of explanations for those who don’t like the idea of octopuses seeing into the future.  The killjoys.

Weak correlation between article quality and peer-review scores

If you, like me, have just had a paper rejected (actually, I had two rejections in one week!), here’s something to cheer you up. A paper in PLoS One by Bornmann and Daniel investigates the relationship between reviewers’ scores of papers submitted to Angewandte Chemie International Edition (one of the top chemistry journals) and the eventual numbers of citations the articles receive upon publication in the Angewandte or elsewhere. Here’s one of their plots:

Useless referees?

Useless referees?

The authors do some regression, and it looks like there is a significant relationship between referees’ scores and eventual impact of the paper, but Jings n Crivens! what a weak relationship! The (Spearman’s) correlation coefficient is only 0.17! If you just eyeball the means of the impact of accepted and rejected papers, it doesn’t make the referees look that good, eh?

I’m really curious about how this would extend to stats or epi journals… but not curious enough to spend time finding out.