If there was anything that I took from my first year psych course in college (I went on to get a degree in it), it was my professor's favorite saying: "Correlation does not imply causation."

The example that inevitably went with that phrase was this: "Ice cream sales are highly correlated with drowning." Now, any reasonable person will see that sentance, and correctly surmise that, although highly correlated, buying ice cream does not cause people to drown. However, buying ice cream and going in the water are two common activities during the summer, thus the rates at which ice cream is sold and the rate at which people drown tends to happen at the same time: summer.

I keep seeing poor research trying to create a cause and effect arguement out of a strong correlation.

From Salon.com, part of the "reparative therapy" series, comes this gem:

There is a high correlation between poor eye-hand coordination and same-sex attraction.

First of all, I'd want someone to prove that correlation with reliable figures before I buy into it. Secondly, like the ice cream and drowning example, poor eye-hand coordination and homosexuality might both be linked to a third factor - perhaps both are genetic. Or perhaps the correlation is best flipped upside down to look for a cause - could homosexuality cause people to avoid activities commonly thought to improve eye-hand coordination, such as team sports? My personal opinion is that the statement itself is crap, and if based on real studies at all, studies where the effiminate stereotype of gay men was not a blind factor. In other words, if the statement was based on a study, I suspect that it was a study where the researchers had expectations of what they would find and failed to design the study to test those expectations.

This study, which hit the news about a month ago, is another example:

Academics Elena Bardasi and Mark Taylor found that a married man whose wife does not go out to work but is primarily responsible for the cooking and cleaning earns about 3 percent more than comparably employed single men.

This study irritates me to no end. In most of the news coverage it has received, there is very little mentioned of possible causes, just an acceptance that the cause is obviously the correlation itself.

When I first blogged about this study, I listed a set of possible reasons why this might be. Here they are:

• Married men whose wives do not perform paid work need to aggressively pursue higher paying jobs in order to generate household income. Married men with another income are freed to pursue careers that interest them instead of pursuing only a paycheck.

• Married men supporting a stay-home wife may be given more promotions by managers who recognize that they need the extra income.
• Married men whose wives also work are less able to move or travel for a better job, due to an increased need for their abilities at home.
• Married men whose wives do not perform paid work have more time available to put in overtime, thus increasing their gross pay.
• Married men who already have a higher income level may be more able to support a stay-home wife.
  Factors such as industry, type of career, amount of education, economic class, etc, may not have been accounted for. If I were still a student, I might well get a copy of the full report and review it to see if factors such as this were accounted for.

Maybe I'll start a foundation for resisting the effects of bad research....