Saturday, May 31, 2014

This to That

If you like to play around with numbers, some might call you a numbers geek. If you like to play around with statistics, some might call you a stats nerd. But if you like to play around with how things do or don't relate to one another, or the difference between causation and correlation, then they would call you Tyler Vigen.

Homeschooled by his mom, Tyler found that in spite of his love for science and math, they weren't areas at which he was particularly good. That didn't deter from studying, however; he has an AA in Liberal Arts from Normandale Community College in Minnesota, an AAS in Intelligence Operations from Cochise College in Arizona, a BA in Criminal Justice from Metropolitan State University in Minnesota, and is currently studying for his JD at Harvard. Still, in spite of his interest in the law, "if someone offered me a job sweeping up the dirt underneath the Large Hadron Collider at CERN, I would very seriously consider taking it."

So what's a boy with that interest to do when he gets bored studying the intricacies of restraining orders? Most people might sit and watch a rerun of X-Men, or play a few rounds of Call of Duty. Tyler, he creates a website. ("Where did I learn to code? My mom sent me to the library and said, ‘learn how to make a website.' So I did!") And since he has a particular interest in statistics and numerical research, he decided to focus on correlation and causation.

Now, if you consider balancing your checkbook on a par with particle physics, your eyes might be rolling up in your head about now. But think of it this way: every day you get up and the "Today" show is on. Also, every day the sun comes up. One doesn't happen without the other. So does the sun cause "Today" to happen? Is it the other way around? Or what?

It's a scientific challenge that has been around literally forever. There are lots of things out there, and lots of other things that are somehow related. The question is does one make the other happen, or is it just coincidence? Does smoking cause lung cancer? Do people cause climate change? Was disco responsible for the demise of records? Or maybe for us landing on the moon? After all, both happened at the same time.

And that's the point. Just because two things happen together doesn't mean one caused the other. To prove the point, Tyler decided to take several sets of data, and map one against the other. The result is his website "Spurious Correlations." At last count it showed 23,299 examples where one set of data is compared to another. For instance, using number sets from 1999 to 2010, he shows that there is a very high correlation between the number of people who were electrocuted by power lines, and the marriage rate in Alabama. Likewise for suicides by hanging, strangulation and suffocation, and the number of lawyers in North Carolina. Does one cause the other? Well, does it?

Tyler puts it this way. He notes that the divorce rate in Maine somewhat matches the per capita consumption of margarine in the US, while that the number of worldwide space launches lines up with the number of sociology doctorate awarded. "Everyone knows there is no connection between margarine and marriage. But everyone considered the hypothesis for a split second, and said ‘are we launching sociologists into space?' And that's the part that's so cool! We were scientists, if only for a moment." And that's what we wants people to realize:  while research is about discovery, and computers can help, only humans can make the decisions and distinctions that advance our understanding.

Tyler said that he hopes that people think more about connections, and which are real and which are not. "I also hope," he says, "that people are less inclined to believe sensationalist headlines where researchers ‘find a connection' between two things." As for now, he's just having fun with the numbers. And while the site has created a bit of a buzz, with links appearing multiple places and even being translated into Russian, he notes that the best reaction he has gotten so far was that "someone emailed me specifically to say that I need to change my hairstyle." The big question I have: what does THAT correlate to?

-END-

Marc Wollin of Bedford has always liked numbers.
Tyler's connections correlate to http://www.tylervigen.com/. 
Marc's column appears regularly in The Record-Review, The Scarsdale Inquirer and online at http://www.glancingaskance.blogspot.com/, as well as via Facebook, LinkedIn and Twitter.

1 comment:

Todd said...

Great essay. In our society where the influence of evidence-deniers is growing, I'm glad to see this example of faulty logic being made so easy to understand. And the causation / correlation misuse promises to get much worse with Big Data.