Voting, working

My voting today was a bit damper than last year when I wrote this sunny little essay. I'm willing to share that I voted for AV. But then I am a statistician, and most people are not. Something tells me that we will not be adopting this system any time soon.

Today the company I work for was taken over. We are assured that it's only a takeover on paper and is actually a great opportunity, and that may be right. It can be so hard to tell with these things. I've been through quite a few of these big changes in a previous job and was largely the same at the end of most of them. I've consequently become quite stoic about the whole thing.

Spot on

Once again, the wonderful xkcd hits the nail on the head. It's actually not bad as an explanation of Type I errors.

The Diet Delusion

I recently finished Gary Taubes' excellent book The Diet Delusion. The first thing to note is that the title misleads, as the book is only partly about diets and is most certainly not a diet book. (It's no doubt an attempt by the publishers to cash in on The God Delusion, but there's no resemblance to Richard Dawkins that I could see). In the US, the book's title was Good Calories, Bad Calories, which gets much closer to its subject. And--unusually for an American book--the cover is better.

TDD argues that much of what we accept as true about diet and disease is actually little more than a set of plausible assertions that were accepted about 30-40 years ago, but that may in fact be totally untrue. Saturated fat may not actually be bad for you. Excessive carbohydrate might be the culprit instead. Eating fat does not make you fat. Excess calories do not necessarily make you fat. Exercise is not normally an aid to weight loss, and often has the opposite effect, as it boosts appetite. Refined carbohydrates make you gain weight by raising insulin levels, which then stores the food as fat. The nub of it may be that, unfortunately, we still have the same design spec as our Palaeolithic forebears, and they just weren't set up to cope with sugar and starch rather than meat, nuts, and berries.

If you want a more extensive summary, see the ten-point conclusion quoted in the review on David Colquhoun's website, where I first heard about TDD.

One of the book's best features is the absence, not only of diet advice, but also of any dogmatic conclusion. Books that put forward a challenging set of ideas can turn shrill and whiny, but that never happens here. Taubes wisely stays above the fray, instead giving us a detailed and well researched history of how our views about what we should eat developed over the last century or so, and how our health may or may not have been influenced by this. One over-arching conclusion is that pinning down cause and effect in such matters is enormously difficult, expensive, and lengthy.

Taubes' narrative shows how many conclusions reached between about 1930 and 1960 were overturned by a number of well-intentioned but forceful individuals who "knew they were right" even when studies failed to back them up. They also didn't read German much: German and Austrian medicine in the 30s had sorted out fat metabolism and diet to a surprising extent, but who was going to turn to these sources in the late 40s?.

This is one of the best factual books I've read in a long while, perhaps comparing with Richard Rhodes history of the Atomic Bomb. I can't say the sorry tale of poor policy-making surprises me. My (admittedly short) experience working in the civil service gave me plenty of examples of policy based on not-very-much. It would also be interesting to hear how UK policies on diet and health fell in behind the USA's lead.

Time to go and eat something fatty, I think.

I'm hungry now

About a year ago, there were various reports in the media of a report from the World Cancer Research Fund that claimed cooked meats raised your risk of bowel cancer. On reading this, of course, my cynicism circuits kicked in and I started searching for the flaw that is usually present in such reports. But I couldn't find one. So maybe I should give up Lorne sausage; a depressing thought. However, David Coloquon has excelled himself with an article about randomised controlled trials, and why dead pig wrapped in bread probably won't kill you.

Good read

I'm currently enjoying Black Swans, by Nassem Nicholas Taleb. He caters to a number of my prejudices, which is always pleasing. He writes entertainingly too. And he's probably right, despite being a bit too rude about statisticians.

Better data

I nearly bought a house today. To be precise: I nearly bought the flat across the landing, but my offer was bettered by somebody else. It's actually quite a promising thing though: I thought my offer would be much too low, but it only missed (apparently) by a small amount. The solicitor was very pleased, as it gave them some idea of what flats like mine are worth (less than they estimated a few days ago). So they now have better data, albeit with a sample size of 1.

Work again

I have acquired a job. No, really this time. It is with the people who didn't quite offer me the last job, but is a different job. Doing proper stats, I hope. My four months or so of semi-idleness is at an end.

In other news, I had a pleasant shopping expedition yesterday to Tiso's. Not only did I get an unexpected discount, but I clocked Iain Banks talking to an assistant.

Is our climate changing?

I enjoy reading some of the more skeptical blogs about global warming. Especially on a day like today, when I'm meant to be doing some real work (i.e. work that earns me money).

I enjoyed the survey of historical weather scares in Anthony Watt's blog. Particularly the global cooling scare from 1975. A lot of these may be to do with the nicely named Pacific Decadal Oscillation. It's all a salutary reminder that the world is Always More Complicated Than That. And it doesn't mean that global warming isn't real, just that you might have to look harder.

Words

I got to use the word "polytomous" today. It makes being a statistician almost worthwhile. I'm still not sure where to put the stress when pronouncing it though (my use was in an email). My paper dictionary doesn't include it, and it baffles the spell-checker.

I'm still working on an excuse to use heteroscedastic, or leptocurtic.

Good for my soul

I don't often do any data entry, but I am at the moment. It makes you realise both what boring work it is, and how you don't really understand your data unless you've got personal with scribbled-on bits of paper.

It becomes clear, for example, that some respondents in a block of multiple choice will fill in all of the first boxes. Or, just to break the monotony, will let the responses zig-zag down the page from one extreme choice to the other. I suppose you could exclude some obvious piss-takers like this, but how many other misleading responses would you still take as valid?

Ask Mr Tufte

My current favourite time-wasting site is that of Edward Tufte, who has narrowly edged out the Exile recently. If you don't know, Mr Tufte is something of a guru in the matter of presenting information, so I suppose it's almost relevant to, but more interesting than, work.

His discussion boards are diverse and intoxicating, taking in radar maps of migrating birds, color charts for farmed salmon, and why PowerPoint is crap (this pleases me, because I think it is).

Right inheritance

I'm currently enjoying Joan Fisher Box's biography of her dad, the statistician R.A. Fisher. It's impossible to study stats without getting to know something about him, but the full story is even more interesting. One surprise for modern readers is his activity in the field of eugenics. At this point we all shuffle nervously, because as modern liberal citizens we don't talk about that embarrassing stuff any more. Fisher talked about it a lot though. He served on a Royal Commission in the early 30s on the problem of the number of mental patients, which much exercised the government of the day. Their final report suggested a program of sterilisation (albeit voluntary). That was about as close as we got in this country to fully fledged eugenic laws. It was ignored by the government and that was that.

Reading the wikipedia article on eugenics filled in some fascinating detail. I liked the Nazi justification of their eugenic policies by the great success that many US states were already having with their own version.

Funny reading

I'm not exactly a great fan of the BMC's membership magazine, which I was browsing on the train, but I didn't expect to find an article mentioning peak oil. There was even a picture of Hubbert's graph from 1956. You should alway treat fifty-year-old graphics with suspicion. The caption says

"The peak arrived right on schedule in 1970. The peak in global oil production is widely expected between 2010 and 2010."

Yes, those sentences do contradict each other. And just in case you're wondering, world oil production didn't peak in 1970. The article is really a thinly disguised puff for a book by David Strahan.

Computers are fast these days

I created a SAS data set with 144 million observations today. Admittedly, it took a while, but it is remarkable that it happened at all.

I, a statistician

I still have some trouble self-identifying as a statistician. It's one of those archetypally dull things, like chartered accountancy--though, thankfully, much more interesting in reality. I've just been doing some tutoring. Hearing yourself say things like:

"No, remember that the null hypothesis states the variances are equal"

and expecting to be understood, is very spooky. And who came up with hypothesis testing, with all its Godawful multiple negatives? Not to mention the frequently misinterpreted p-value.

Is this a way for sane people to talk?

Statistics and owls

Went to an interesting RSS talk last night. The location of these talks is a rather lovely Georgian town house in the New Town. There are quite a lot of portraits of famous mathematicians and scientists on the walls, but I've never noticed before the rather lovely pencil (?) drawing by William Dyce of James Clerk Maxwell as a child, apparently embracing a large owl. I find it a rather compelling image: the young scientist embraces wisdom. Or am I just overloading it, and it's a piece of pure whimsy? I'd love to know more. My main knowledge of Dyce is as the painter of a version of Christ in the wilderness, where the landscape used is very clearly Scottish.