Wednesday, November 24, 2004

What You Do When You Make a Mistake

This seems to depend on who you are. If you are the Center for Disease Control and Prevention you can publish a study in March and come back in November with:"Oh, by the way, our figures in that March study were wrong because of some statistical glitches. And no, we won't tell you how wrong they were. Bye bye!" Nice work if you can get it.

The specific question was about the number of deaths in the year 2000 that could be attributed to obesity, and the March publication gave them as 4,000. Now it seems that the real number is something less, but how much less? Well, we will be told when the Center for Disease Control and Prevention is ready to do so. What did you think they were? Public servants?

I have noticed a lot of carelessness in some medical studies which analyze diseases where a kind of a societal condemnation plays a role. A 1990's study about the deaths caused by cigarette smoking overestimated such deaths quite wildly and illogically by assuming that every person who smoked and died from diseases such as stroke and heart disease would not have died from these causes if she or he had not smoked.

Nobody thinks it's very serious to err in the direction of 'right results', especially when everybody agrees on the issues to begin with. Well, I think that this is when the researchers should be especially careful with their work. Common bias is still a bias.