Reply #11: A statistical perspective

[spooky3]

This is a topic on which there is a lot of research conducted by people in a variety of fields. I'm not knowledgeable enough about it but it always bothers me when people like Lawrence Summers ignore the fact that it is there and draw conclusions without reading it. That's like talking about whether a certain food causes cancer without reading any of the literature testing the hypothesis in a controlled way, maybe based on what happened in his or her family. It's possible the speaker's right but certainly it is far better to read the controlled research than to draw conclusions on the basis of anecdotal experience.

I do know this dangerously small amount:

1. from a statistical perspective, on many of the characteristics on which men and women have been compared, the means are not different to a statistically reliable ("significant") standard. And given that journals strongly prefer articles where significant differences are found (because there are infinite numbers of reasons for why one didn't get significant results) that means you're not seeing a lot more studies showing no gender differences even if they were well done.

2. Many studies showing differences have been discredited because of poor research design, so better research needs to be done.

3. On those good studies where differences are found, despite mean differences, there is considerable overlap in the 2 distributions. Picture one bell curve a little to the right of another. What this shows is that while it's correct to say men are/have more of something than women (or vice versa) there is a lot of variance in the distributions unexplained by gender, such that there are many women who score higher than many men. Under these circumstances it would be wrong to say men and women as groups don't differ but it would be just as wrong to treat individuals as though their own score is primarily determined by gender. Obviously other characteristics (for example, intelligence) may be much more important in determining an individual's score. And, even where gender means are different, it's entirely possible that something correlated with gender (such as nurture) could account for the difference.