A Good Example of Statistical Honesty

By in

I was reading this article in the economist

“Sexual Selection” The Economist December 11 issue

 In this article they review work on the selection of a mate for humans.  It was published by one researcher that women, when ranking photos of men with different levels of perceived masculinity, chose the more masculine photos more frequently in countries where safety and violence were big concerns.  The published result was that women linked the masculinity with the aggressiveness and survival skills of manly men.  Their data showed a significant correlation between violence and the selection of manly men.

 Another set of research that ranked photos of men showed that countries that had health concerns chose masculine men because of a perception that their genes would lead to better children and more children in that environment.  Their data showed a significant correlation between the health risk and the selection of manly men.

So how could both be true?  It turns out that most countries that have health concerns also have violence issues.  So since the predictors are correlated, which one is the true causal factor. 

When taking the two factors together and then accounting for the GDP of the country, the violence issue goes away, but the health issue remains predictive.  Here is the quote from the article;

“The statistical tussle shows the difficulty of drawing firm conclusions from correlations alone. Dr DeBruine and Dr Brooks admit as much, and agree the dispute will not be settled until the factors that shape mating preferences are tested directly.”

The key point is that correlation is not enough to show causation.  Many social science work is done through collecting historical demographic data where they find correlation, and it is disproven by a researcher in a few years.  Be careful.

Enhanced by Zemanta
(0 votes. Average 0 of 5)