tag:blogger.com,1999:blog-4632722460026573847.post4854155680111550525..comments2013-11-29T01:02:04.202-08:00Comments on Sociological Entrepreneur: Why Aren't There More Female Programmers?Kristinahttp://www.blogger.com/profile/14798780312617916693noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-4632722460026573847.post-58821025119082249412013-06-14T16:57:37.989-07:002013-06-14T16:57:37.989-07:00You are making a classic error here when citing
S...You are making a classic error here when citing<br /><br />Sex differences in quantitative and analytical GRE performance: An exploratory study<br /><br />in support of your conclusions.<br /><br />That error is assuming that measures of central tendency (mean, median, etc) are the default when attempting to explain observed differences in groups.<br /><br />With measurements such as GRE analytical score, the distribution of scores is very important to tackling questions such as what you have set out for yourself here. The means of the two groups can be exactly the same yet one group be the majority in a field based upon greater spread in the distribution. That is, if you assume a threshold score above which high achievement in a field is possible, then a more dispersed population will have more above that threshold. In other words, let's say that being a great programmer requires an analytical GRE score above 700. If men are more dispersed, more very high and very low scores on this test, then those above 700 will be a majority men. Think this threw by drawing a few sample bell curves and make the male curve less peaked and with thicker tails, which describes a situation of greater variance.<br /><br />It turns out that on most things men are more varied than women. This includes many tests of mental capacity as well as things like height.<br /><br />I downloaded and skimmed the results of the paper I mentioned above that you referenced. The authors do not give measures for variance of groups, instead they focus on gender effect based upon median scores.<br /><br />Achieving statistical significance in a difference in median or another measure does not address the issue I stated previously related to variance.<br /><br />Skimming the data tables though, I can tentatively predict that their data showed greater variance among men. How can I tell? In the analytical section, the men's mean is higher than that for women, however the women's median is higher than that for men. Also, the men's mean is around 76 points higher than the median while women show around a 38 point difference between the two.<br /><br />This indicates that there are more men at the highest score levels and this pulls their mean upward but would not influence the median. The mean's median is in fact lower.<br /><br />How did I figure this out and you didn't? Well, either it is some built in bias or I can thank those thick tails.DubVhttps://www.blogger.com/profile/14367764785422714196noreply@blogger.com