This course will become read-only in the near future. Tell us at community.p2pu.org if that is a problem.

Lurking Variables


 

Dilbert.com

In plain english, correlation is used to tell us that two things are related, and usually implies that there is a causal link. Unfortunately, as a piece of statistical jargon correlation means something else. It just tells us that there is a linear association.  

I don't really want to spend time explaining more than that. We need to think about lurking variables.

If I said weekly gas sales were strongly correlated with weekly counts of elderly deaths what would you say? Elderly people are having gas accidents?  

It's probably very obvious, but in cold weeks lots of gas gets sold, in cold weeks, lots of old people die. We call the weather a lurking variable, which explains the apparent (circumstantial) association between weekly gas sales and weekly elderly deaths; but which may (probably) not actually explain the correlation.

 

 

Task: One of the articles in the Headline Database concerns a report by Sheila's Wheels, an insurance company with (I guess, given their name) a fairly well focused target audience. It claims that females wearing short skirts in the summer causes male drivers to have accidents.

Please comment on whether you think this is possible. Do you think there is a direct link Could there be any lurking variables?

Task Discussion