Errata.

When I first posted about the proportion of women in the facebook group "I Love the Smell of Econometrics in the Morning", I hadn't properly covered hypothesis testing in statistics, and so had only a heuristic understanding. I'm sorry to say that I got it quite wrong. I'm a little embarrassed. Also, the fact that nobody called me on it testifies either to the apathy or statistical innumeracy of my readership. Or possibly the absence of any readership at all. Anyway, we have more data, and I have corrected my analysis below.

We will use a most powerful likelihood ratio test to distinguish between the composite hypotheses H0: p>p0 and H1: pp0, is decreasing in x and p0, so the uniformly most powerful test at 90% confidence for our hypothesis is to reject H0 if x=k)>=0.9.

New members have joined since last time, so we now have a larger sample in which we observe fourteen girls out of a total forty-eight members. If p0=0.4, then k=15, so our sample allows us to reject the hypothesis that econometricians are, on average, more than 40% girls with 90% confidence.

This result is less striking than last time, where I (using pretty erroneous statistics) asserted that the number was less than 35%. As a dude, I have to say that this revision is a good thing.