Retorts

Predicting Gender

The Dominion (14 June 1999) refers to a new study that shows women with high testosterone levels (barristers and solicitors) are more likely to give birth to boys than to girls. The study, from the American Psychological Society, involved 58 barristers, 66 solicitors and 16 beauty contestants.

Assuming that each subject had only one live birth each during the study, barristers had a 60 to 40 chance of having a boy. I guess, to the best of my ability, that this was 35 boy and 23 girl babies. The null hypothesis should be that boys' and girls' births are naturally even, and this should be reasonably easy to test statistically, i.e. scientifically.

Well, let's do it! The expected number of boys is 58 x .5 equals 29. The variance is npq = 58x.5x.5 equals 14.5. So one standard deviation is sigma equals 4. The observed data does not even ring a bell at the 5 per cent significance level (binomial distribution model). A 95% (2s) confidence interval is 29-8 = 21 to 29+8 = 37. Thirty-five boys is within this confidence interval.

I predict that the American Psychological Society is out of its mind and will fold. Also, the Dominion has not verified the reasonableness of its assertion "Gender goes with the job".

Those barristers could persuade the judge anything they like.

Don McDonald, Wellington