Reader Comments

Post a new comment on this article

Problems with statistics and design

Posted by samwang on 05 Feb 2012 at 16:25 GMT

Commenter jshyde has outlined some of the problems. First, the combination of many small group differences using Mahalanobis distance is misleading, and would lead to a large apparent difference between any two relatively similar entities - Serbs and Bosnians, locations in North and South Dakota, what have you. Second, the personality measures are self-reported, which leads to some question of whether the answers carry a small bias driven by cultural expectations. Even a small bias would appear large if added according to the rather unorthodox statistical approach. In addition, these findings seem to claim considerably larger gender differences than previous research.

No competing interests declared.