Follow-up on women in science.... A few weeks ago when the whole sexism and the blogosphere stuff was being floated, Susanna of bias brought up the study alluded to in the text. I was curious as its result seemed so striking. I asked a friend, who I will call B, to check it out. Here is a summary of what B had to say.... B's text (my emphasis): The article by Stewart Lee is a restatement with a little cheerleading of the "commentary" article in Nature (Wenneras and Wold, Nature, vol 387, p341-343 - sorry it's not available online even for those with online access - I think this is because it's a commentary article). These articles attempt to quantify gender-based discrimination in the awarding of biomedical research funds to post-doctoral candidates in Sweden. Surprise! They find discrimination and conclude that women have to be 2.5 times as productive as average male to obtain the same level of perceived competence. But this is simply absurd. There are a miniscule (relative to the total number funded) number of individuals that are 2.5 more "productive" then the average grad student/post doc - there is just nowhere near that kind of variance in the group. It basically says the bell curves describing the perceived competence of men and women have virtually no overlap. This simply is not true. Their made up model gives made up numbers. It's a problem that has to be modeled because you need some way to map numbers to these ill-defined concepts, like "perceived competence" and "unfair discrimination" and the various factors that contribute to these things. For complex issues, like subtle forms of discrimination, there is no "correct" way to model the system and the modeling just yields the scientific sounding version of the opinion of the authors. The model used by Wenneras and Wold illustrates this. They constructed their model in a plausible and rational way, trying to take into account lots of relevant factors and to quantify things "scientifically." They determined from studies that non-proportional representation in funding stemmed from lower perceived scientific competence of women and modeled this as function of publication productivity. They arbitrarily defined productivity in terms of "impact points", which are a measure of the quantity of publications and reputation of journals they are published in. They arbitrarily gave publications like Nature and Science 10 impact points per article while more specialized publications receive 2-3 impact points per article. They said they took into account a variety of complicated factors: prestige of university/research advisor, priority of research area, first authorship, difference in degrees (there were medical doctors, nurses, and PhDs - 59% of the men and 26% of the women were medical doctors, with a lot of that difference being made up by nurses). But these things are again pretty difficult to quantify, though this was necessary, since they used multi-dimensional linear regression analysis to make all the interdependency of these variables go away. I might also question where the perceived competence should be linear, as the difference in perceived competence between no publications and 1 publication is probably a lot larger then the difference between 10 and 11 publications. And they are basing all of this on a sample size of 114 people, 20 of whom were funded. The breakdown was as follows: women (4/52/8%) and men (16/62/25%) - (funded applicants/total applicants/ percentage). In addition, the authors determined that women start 30 impact points behind men in perceived competence. The authors of the Nature article and the author of the Natural Science Review explicitly state that women are the equivalent of 3 Nature or Science articles or 20-30 journal articles behind men. The article also notes that 3 Science/Nature articles are the number many researchers are happy to get in a career, and I agree with this. So then for the authors elaborate model to hold, in a ~10 years of graduate and post doctoral work a female has to publish 20-30 more papers or publish 3 more Science/Nature articles than men. But 20-30 articles are already on the high end of what anyone does over that time frame. The model implies it is virtually impossible for a women to be perceived as competent as the average man, but women do become professors and do receive funding. In fact the authors note that of for some do get funded.... Razib adds: I'll do my own digging when I have time. All I can say is that the results were so "good" from the perspective of someone trying to show blatant discrimination that I was astounded, and thought it was either some serious data-massaging or my world was going to be rocked. I definately lean toward the position of B, and tend to believe that discrimination levels this high, if they hold up, can be chalked up to the small sample size, and the fact that it was restricted to Sweden (yes, yes, I'm sure Sweden is way more progressive than the United States....).