Discussion 3 to Ask the Patriarch 15
On Correlation
by: Cletus
To add to this discussion (or any other,) please use the Contact form. This exchange of views has been continued.
I just wanted to comment on the statement about a correlation between IQ and intelligence.
Correlation is a very abused statistical tool. A correlation in and of itself is indicative of absolutely nothing besides having a correlation.
I could get a positive correlation by looking at "ice cream sales" vs "shark attack". It would tell me that when ice cream sales are more frequent then so are shark attacks. You'd be hard pressed to tell me anything remotely causal was established by the data/correlation. Here the hidden variables "time of year" or perhaps "temperature" are evident.
But when things become more complicated things often get clouded. I could even have an R2 of 1 and it could still mean nothing. So I guess my take home message is before you leap in and start calculating correlations be sure you're aware of what it is you're actually calculating. A LOT of people screw this up and conclude that ice cream sales causes shark attack. So heads up.