Sunday, May 16, 2010

The Difference Between Correlation and Causation is Statistically Insignificant

Read this News in Brief from America's Finest News Source.

Clearly, the examples the The Onion gives are ridiculous.  There are plenty of times, though, where real people try to find meaningful connections between data that is simply random correlation. 

A cool example where the correlation is meaningful, though, is the orange market.  The companies that interact in the orange commodities market care a lot about the weather in Florida.  If there is a frost in Florida, oranges die, orange supplies go down, and the price of oranges go up.  These companies generally aren't satisfied with the detail and quality of the weather forecast provided by typical sources, so they hire their own meteorologists to hang out in Florida during the colder months to give private detailed reports.  They then make trades in the orange futures market based on the data they receive.  The result is that when there is a chance of frost during a few days or weeks, the orange futures market is a better predictor of actual frost than the National Weather Service.  If orange futures spike up in price during the afternoon, expect frost that night.

For another post on interpreting data better, read Points are for Sports, Pay Attention to Percentage in Finance.

No comments:

Post a Comment