Friday, October 22, 2010

Why Correlation Is Not Causation

I've been thinking a lot about regrets lately, mainly in terms of my college choice.  I keep going back and analyzing what I was thinking at the time, how I could have ever thought that I would be happy with the choice I made.  A lot of people choose a school for its academic programs, but my choice was based entirely on where I would have the most fun, be the happiest, and feel like I belong.  With those priorities, it's hard to understand what went wrong.

Then I starting thinking about my psych stats class, which psychology majors were required to take.  Most students found this class a bit of a drag, but there were parts of the class that I had fun with, particularly correlation without causation.  For example: if a study shows that students who sit in the front of the class tend to do better in school than students who sit in the back, you can't tell for sure whether sitting up front causes students to do better, because there could be other reasons for the correlation (ex: maybe students who are more studious tend to sit in the front).  In stats class, we were often given homework problems in which someone would claim that A causes B, and we would have to explain why this wasn't necessarily true and come up with alternative explanations.  I loved thinking up alternative explanations.  There are only three basic types of correlations: A causes B, B causes A, or a third factor C causes both A and B.  But since that third factor could be anything, the explanations can be infinite.

But for a required class that was only supposed to teach us how to run a psych study, I realize now that it's really applicable to other things. Here are a couple of statements that might have appeared on our homework, statements we would be asked to challenge:

1. Students at school A have higher scores on a large number of positive, socially desirable things than students at school B.  Therefore, school A is better for everyone.

2. A student wants to go to the college where she'll have the most fun, be the happiest, and feel like she belongs.  The best college for her would be a place where students have the most fun, are the happiest, and feel a sense of belonging

It's obvious to me that these statements are not true, and I could list a ton of reasons why with alternative explanations.  But I didn't always get it.  There was a time when I believed these statements, back when I applied to college. Maybe I needed all those examples in statistics class to really understand.  I hope that I'll be less naive now, that I will no longer look at things and see the explanation that I want to see, when a million other explanations are possible.  And if I ever help someone look at colleges or jobs or anything, I will use my understanding of correlation explanations to help them make the right personal choice.

No comments:

Post a Comment