Tuesday, September 26, 2006
The Correlation Confusion
The LSAT is this weekend, and for those who are about to undergo the three hour ordeal, I'm wishing you the very best of luck. As you go into the exam, be sure to look out for one common area of confusion that bedevils everyone from nervous test takers to high-level policy-makers, and that is the difference between "correlation" and "causation."
The source of the confusion stems from the fact that it is relatively easy to observe when two events tend to go together, but it is relatively hard to tell whether one event is causing the other. Two events are "correlated" if the presence of one event means that the presence of the other event is more likely. A causal relationship, however, only exists if one of the events is actually causing the other. Or as one statistician puts it, "correlation does not imply causation."
For example, someone might observe that, in the summer, drownings at the beach increase and so do sales of soft drinks. Therefore, there is a correlation between drownings and soft drink sales. But it does not necessarily follow that drownings cause people to buy soft drinks or that people buying soft drinks cause drownings.
Nevertheless, if two events tend to go together, there often is an almost irresistible impulse to conclude that one causes the other. On the LSAT (and in life) before you accept someone's conclusion that event X is causing event Y just because the two events are correlated, you should first consider these possibilities:
Only if you have considered and found reasons to dismiss these other possibilities should you be ready to accept that a correlation might demonstrate a causal relationship.
Well, that's my pre-LSAT tip for the day. Good luck, test takers!
The source of the confusion stems from the fact that it is relatively easy to observe when two events tend to go together, but it is relatively hard to tell whether one event is causing the other. Two events are "correlated" if the presence of one event means that the presence of the other event is more likely. A causal relationship, however, only exists if one of the events is actually causing the other. Or as one statistician puts it, "correlation does not imply causation."
For example, someone might observe that, in the summer, drownings at the beach increase and so do sales of soft drinks. Therefore, there is a correlation between drownings and soft drink sales. But it does not necessarily follow that drownings cause people to buy soft drinks or that people buying soft drinks cause drownings.
Nevertheless, if two events tend to go together, there often is an almost irresistible impulse to conclude that one causes the other. On the LSAT (and in life) before you accept someone's conclusion that event X is causing event Y just because the two events are correlated, you should first consider these possibilities:
1. Instead of X causing Y, maybe Y is really causing X.
2. Maybe a third event, Z, that hasn't been considered yet is causing both X and Y.
3. Maybe there is no causal relationship at all.
Only if you have considered and found reasons to dismiss these other possibilities should you be ready to accept that a correlation might demonstrate a causal relationship.
Well, that's my pre-LSAT tip for the day. Good luck, test takers!
