One of the most common human error is to confuse a causal effect with a correlation.
A causal effect, as the name implies, means that one of the items being considered has a mechanism for producing some kind of change or action in the other. Thus, if you have rhinovirus organisms present around you, you're very likely to catch cold because one will cause the other.
A correlation is where items are simply statistically linked. There may be an underlying causal mechanism -- this is usually where they're first considered and may eventually be discovered, but there's no proof that one actually causes the other. There is simply a close association. An example of *this* might be being cold and wet and then catching cold. It's often correlated, but simply being cold and wet won't make you sick in any direct fashion.
There are a lot of examples of this (correlation of the stock market with skirt lengths or sidewalk temperatures, for instance). Most of these are pretty silly, but using this natural tendency to link two closely associated things as one causing the other is the basis of many human undertakings from superstitions to nutrition and advertising.
Where is all this going?
Right here: http://statestats.appspot.com/
You can look at the correlation between search terms on Google, and the statistics about people the states that are doing these searches. Thus, if you search for "cats", you'll find that there's 100% correlation with people from New Hampshire, and there will be a column of related statistics having to do with people from those states (obesity, education, same-sex marriage etc.) that is hilarious. You'll find yourself jumping to conclusions right away.
You can go wild with this - it doesn't flinch from nasty search terms so you can see where the pervs live (or at least correlate well). Also, you can link search terms with a "+" sign. Try "mittens+sex" -- Texas and Kansas are the winners there.
It makes one wonder.
Do read the "Be careful drawing conclusions from this data" disclaimer.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment