Today, I'm debuting a series of interviews called "Numbersense Pros". These are profiles of people who I turn to for well-reasoned data analyses. If you follow this blog, or read my new book, you'll understand why I say "well-reasoned" as opposed to "correct" or "true". The best way to develop your own Numbersense is to learn from others. In the interviews, I ask how they learned to analyze data, and what they read for inspiration.
The first interviewee is Felix Salmon, currently the finance blogger for Reuters. You can see his blog here. He comments on mostly economics and finance related matters, and sometimes makes funny videos (for example, here). I am a big fan of his writing.
***
KF: How did you pick up your impressive statistical reasoning skills?
FS:
The first cousin of statistics is probability, which I learned from being taught backgammon at an early age. Then there's this thing called Further Applied Mathematics which is studied at high school in the UK but which is pretty college-level stuff by US standards. That's about it, really.
KF: If I may push you a little further, I'd say most of the people who took Math at school do not develop the ability to judge data in real life; what differentiates you from the crowd?
FS:
The Maths A-levels were split into Pure and Applied. I was reasonably good at both, but I just breezed through the probability and statistics bit of Applied -- I found it by far the easiest part of the whole shebang. To put it another way: it's not something I learned, it's something I've always had an intuitive feel for. And I'm good at smelling bad numbers. When I'm proof-reading or editing somebody else's work, or when I'm reading something on the internet, I can just *tell* when a number is wrong. A few days ago, at a conference, a woman got up on stage and said "you can now fit all the world's music on a single $600 disk drive". And I just *knew*, without looking anything up, that it wasn't true. Don't ask me how.
FS:
The data I encounter most often is market data, where my biggest pet peeve is that no one ever stops to think whether they're looking at signal or noise. Specifically, market journalism seems obsessed with one-day moves, even though they (ought to) know that one-day moves are nearly always noise rather than signal.
FS:
The iterative blogosphere. Any one blog post can easily be erroneous. But if you get a real blogospheric conversation going, the end result is likely to be pretty robust and sophisticated.
KF: Thank you.
In Australia we are obsessed with one month changes in our unemployment figures. They are generally about plus or minus 0.1 percentage point. Unfortunately the 95% CI on changes is plus or minus 0.2, so just noise. Every now and then a journalist who actually understands will write an article explaining that it is all pointless, but nothing seems to change.
I'm also surprised by how often people will confuse millions and billions, especially when it is an odd context. So a company will be described as losing $60 billion a month, when it should be $60 million, because nobody can do the math and work out that $60 billion per month is more like the US federal deficit
Posted by: Ken | 08/05/2013 at 08:42 AM