Related Posts

Share This

Causation vs. Correlation

I am going to an event called Blogger Social in New York in April. One of the attendees, Steve Woodruff has pulled off the ultimate link bait and is doing profiles of the attendees so that you can learn about who is going to be there, before you get to the event. As a person with a last name that starts with “W” this is one of the few times I am getting lucky, I should be up there close to the event date and after the whole world is checking his site (assuming of course he makes it through the list before dying from exhaustion). This is a great example of social media in action, it’s quite possible that the 70 attendees at the event can get to know each other to some degree, even before attending the event.

Today featured Matthew Bailey, who has done a great piece of research on red shirted crewmen in Star Trek, and uses it as an example of  analytics in action. He also cites Edward Tufte, probably the closest thing to a rock star that you can find in the world of graphing (and by extension, economics).

There were two things the red shirt analysis brought to light for me. One is the recurring theme in marketing of testing raising more questions than answers. I find it very common to set up a test and have it raise additional questions that weren’t even considered in the first round. The other is “causation vs. correlation”, I think the last discussion I read on that was in Freakonomics, the fact that characteristics that a group of individuals have in common is not necessarily the reason they are grouped together. For example: yes, there were many red shirted crewmen that died, but that’s because the red shirt signifies the security team, people put in dangerous situations. There’s nothing inherently dangerous about a red shirt versus a blue one.

When causation is confused with correlation, this can lead to problems. In the past I have worked at organizations that would look at this data, ship some blue shirts to the security team, and declare victory.