One of the common mistakes that non-scientists make is the classic correlation vs. causation conflation. You know, when they find that A and B show some sort of correlation and then leap to the AHA! moment and incorrectly imply that B is caused by A.

Correlation is when two data sets, A and B, track each other; the closer they track, the higher they correlate. Statisticians actually have a number to measure this tracking: the correlation coefficient. The value can be between -1 and 1 such that perfectly tracking (as A rises, B rises) data have a correlation coefficient of 1 and perfectly opposite data (as A rises B falls) have a correlation coefficient of -1.

Causation is how you get from simple correlation to cause and effect. For example if you drop different masses, m, you can measure the measure the force, F, at which they hit the ground. Correlation shows that as m increases, F increases* and nothing more*. In other words, correlation is a black box; causation is the model of how to get from m to F.

We all know that storks bring babies. And we know that is an old wife’s tale. It was illustrated back in the Victorian Era when statistics was beginning to become a real branch of mathematics. Some wag took the number of stork sightings in London and looked at birth records from the same years. He found that higher birth rates seemed to track with higher number of stork sightings. So he plotted the results and voila! The birth of scientific malpractice was born.

Of course, we know that any two unrelated data sets can be plotted against each other, and over the years this game has been played to supposedly “prove” a point. However, the point they supposedly prove is essentially pointless, as there is no mechanistic explanation as to how the two data sets are related.

This method is how charlatans convince people of their science cred. Because if it is in sciencey looking graph, it must be legit. Which is why polio is making a comeback.

The folks at Spurious Correlations show the ridiculous extremes one can play this game.

### Like this:

Like Loading...

*Related*