Echo and Bounce

Data

Last year, I read a heap of books. Way too many, probably. One big chunk of this reading was what I like to call a “rationality binge”. Spurred by the excellent Harry Potter and the Methods of Rationality and by Straight and Crooked Thinking, I’ve long wanted to get a better grip on rational discourse.

This has been amplified by other books I’ve read that dare to suggest that decisions in companies should be based on evidence, and that you should treat your crazy idea for a new product like a hypothesis that needs falsifying.

So I guess it’s not right to say I want a better grip on rational discourse. What I actually want is to help myself and my friends stop wasting their time on things that sound plausible but are actually utter tosh.

Amongst our weapons are clarity, logic, rhetoric, and an almost fanatical devotion to actual data.

“Data” isn’t in the classical trivium, but is just as essential as the other three and just as tricky.  Humans are really bad at thinking statistically and there is more data out there than ever before.

I’ve been trying to work my way through Think Stats, which came recommended by friends, but I got bogged down trying to work through it in a weird programming language rather than just sticking with the one in the book. Even today, I’m ashamed that I don’t really understand Bayes’ theorem or its consequences, and embarrassed that Nate Silver’s work in the recent U.S. election might as well be magic to me.

Still, I’ve been plugging away at it. I now know the difference between a cross-sectional study and a longitudinal study. I kind of, sort of know what statistical significance is, even though I don’t really have the tools to figure it out. I’ve been able to read papers on the perils of relying on the results of web experiments and actually follow them.

I haven’t yet had much of a chance to apply these things professionally, but my plan is to skill up (i.e. finish reading through Think Stats), and then apply statistical reasoning to my next couple of personal projects.

At the bare bones level of actually gathering data and tracking how it changes, it’s not too much of an exaggeration to say that it has been life changing.

As boring as it sounds, a bunch of numbers in a couple of spreadsheets have been my greatest allies in my constant quest to not be a debt-enslaved fat ass.

Left to my natural impulses, I will buy whatever feels good at the time, eat whatever you put in front of me, and then head out later to buy more food to eat. This isn’t exactly sustainable behaviour. So, for the last little while I’ve been tracking my spending and recording my weight and such. For my fitness goals I’ve got fairly detailed records that go back over three years.

As proof:


Which tells the story that I lost a lot of weight in 2011, have kept most of it off while mostly not caring about weight loss, and have recently started to care again, with okayish results.

Because I have to explicitly do work to take the measurements and record them, I only do so when I care, and more often when I’m feeling good about myself physically. And when I track myself more closely, I tend to be more disciplined about how I eat.  Or at least, that’s what I think. I don’t know how I’d go about inferring it from the data.

One great thing about having this data for so long a period is that I can set reasonable expectations what progress I’ll make over the next few months. This gets me excited – I like working toward concrete goals – and will hopefully spare me disappointment. Best of both worlds.

I realize that all of this is a bit nerdy, but I can’t help but think it’s of fundamental importance. Not me losing weight – that’s just vanity dressed up as good sense – but measuring things that matter to us, and using that data and our reason to make decisions can only improve the world.

But if I, as a dedicated geek with a maths degree, can’t be bothered working through a basic stats intro book, how can I expect others to do so?

Perhaps it’s because I’m afraid that as the tools get better and more data becomes available, a gap will grow between those who are making reality-based decisions and those who are having their decisions made for them.

Next week’s word: exercise.