[2013-12-25] Correlation Equals Causation

[ReasonablyDoubtful]

People seem so quick to bring up the fact that correlation is not the same as causation when they disagree with something, and yet take correlation as causal proof when they're trying to find proof.

[Kaharz]

People seem so quick to bring up the fact that correlation is not the same as causation when they disagree with something, and yet take correlation as causal proof when they're trying to find proof.

Causation is extremely difficult to prove in social science.* There are typically way too many variables and relationships are often complex. So you use correlation. The difference between just pointing to a correlation as proof of causation and actual social science is that you use statistical analysis to determine the strength of the correlation and you test with multiple controls to eliminate the possibility of a spurious relationship. If you are doing everything right, you also don't say, A caused B. You report your statistics regarding the strength of the correlations, the statistical significance of that correlation, and the results of alternative null hypothesis testing. So the best you'll get is something like "In 95% of random samples, A is strongly correlated with B and further testing indicated C, D, E, etc were not intervening variables. So it is likely that A is a contributing cause to B." That is a bit of a over simplification, there are other things you have to shown according to the rules of causation, but that is the basic gist.

*Or really almost any science