Understanding correlation and causation

I wrote the following for another site (now defunct) a few years ago:

You are reading an article in your local paper reporting the findings of some recent research, which reports that the study revealed a link between eating some specific food and an increased risk of cancer.  Should you immediately remove that food from your diet?  How you respond to this information depends on how you interpret the findings, and properly interpreting the findings requires you to understand the relationship between correlation and causation.

Causation occurs when a first event *causes* a second event, while correlation is a broader term that only requires that the two events be related somehow.  While events that have a causal relationship will also be correlated, not every correlation is causal.  For example, the first and second event might *both* be the result of some unmeasured third event.  In the hypothetical example in the first paragraph, the specific food item might be associated with some independent activity that increases the cancer risk, so that merely eliminating the food will have no impact on the cancer risk.

A correlation finding is important in scientific research because it shows that there is some type of relationship between the variables.  These findings help show researchers where to focus further efforts to better understand the exact relationship.  The goal, of course, it to determine causation, but that conclusion can only be reached after carefully evaluating and eliminating the possibility of other types of relationships.

Unfortunately, the popular media tends to blur the lines in reporting these findings.  Where a scientific journal will normally be very specific in identifying the exact nature of the correlation, and will likely point out what further questions need to be answered to determine the nature of the correlation, the popular reporting tends to be a bit sloppy in describing the relationship described in the findings.  Combine this sloppy reporting with a general population that does not have a strong grasp of the relationship between correlation and causation, and people will conclude that the results show a causal relationship, even when that is not the case.  Having an awareness of this relationship allows a reader to analyze such an article critically, in an attempt to discern the exact nature of the relationship discovered.  This in turn, allows the reader to respond to the information in a more appropriate fashion.


I was thinking about this again recently because of all the posts I've been seeing on Facebook about the incident where Tim Tebow prayed over a spectator who was having a seizure.  The implication of these posts is that the seizure stopped *because* of Tebow's prayer.  People see the prayer and they see the seizure end, and they link the two causally.  However, just because the two events are correlated...this does not imply anything about whether one caused the other.  Except in very rare (and dangerous) situations, seizures end.  They are almost always self-limiting.  This persons seizure would have almost certainly have ended, even if Tebow hadn't been there. The end of the seizure can be explained by the medical sciences without any need to reference the prayer. There is no miracle here.  Is Tebow's prayer a genuine gesture of compassion? Only Tebow can say for sure what his motives were, but I'm inclined to take it as such.  Is there value in such gestures?  Certainly, in so far as they provide comfort for those involved and bind us a community.  Did it affect the medical outcome? Not at all.  I'm not opposed to prayer.  By all means, pray for the sick and others in need.  But don't expect those prayers to cure people or help those in need.  Don't *just* pray.  Let your prayers be accompanied by *action*.  DO something.  Get medical attention for the sick.  Lend a hand to those in need.  You need to be the instrument that answers your prayers.