Bad Statistics and the Gender Gap

So there’s an article in Scientific American called How to Close the Gender Gap in the Labo(u)r Force (I’ve added a `u’ to `Labour’ so that it can be understood in the UK).

I was just thinking the other day that it’s been a while since I added any posts to the `Bad Statistics’ folder, but this Scientific American article offers a corker:

That parabola is a  `Regression line’? Seriously? Someone needs to a lesson in how not to over-fit data! It’s plausible that the orange curve might be the best-fitting parabola to the blue points, but that doesn’t mean that it provides a sensible description of the data…

I can see a man walking a dog in the pattern of points to the top right: can I get this observation published in Scientific American?




16 Responses to “Bad Statistics and the Gender Gap”

  1. What is the difference between an economist and an astrologer?

    The astrologer won’t ever tell you that the Sun will rise in the west tomorrow morning, and never set again.

  2. There is also this howler: “During the same period, we experienced the world’s fastest ever decline in fertility.” Have they been putting birth-control pills in the drinking water? Or is something affecting sperm quality?

    What they mean, of course, is that the number of children per woman has gone down. Most people realize that this is due neither to lack of fertility nor to lack of sex (though the latter might be a non-negligible factor contributing to the low birthrate in Japan), but to the fact that more people choose to have fewer children. (Of course, while no-one should be forced to bear children, the idea that a low birthrate (particularly a non-sustainable one) should be used as a proxy for social progress is bizarre.)

    • I’m sure that at least Anton (why not everyone?) will agree with me that the criticism of post-modernism bullshit is more or less on the mark.

      • There is also some other interesting stuff there, this linked from the link above: Most telling, a 2013 paper entitled “Like a Virgin (Mother): Analysis of Data from a Longitudinal, US Population Representative Sample Survey,” published in BMJ reported that 45 of the 7,870 American women studied between 1995 and 2009 said they become pregnant without sex. Who were these immaculately conceiving parthenogenetic Marys? They were twice as likely as other pregnant women to have signed a chastity pledge, and they were significantly more likely to report that their parents had difficulties discussing sex or birth control with them.

      • Anton Garrett Says:

        Bulllshit is too kind a term, because bullshit is daft and wastes time but is not malign.

        I have been predicting for two decades that the new Humanities postmodernists will eventually come after the scientists on campus. The latter mostly do not understand the intensity of the matter and will lose unless they are prepared to fight hard. Meanwhile, it might be best to go to Caltech or MIT or some place of higher education in the sciences that does not have Humanities faculties.

      • telescoper Says:

        We should go back to the good old days when all Humanities professors were Marxists!

      • Anton Garrett Says:

        My more important point is that this is a slow-burn crisis of existential magnitude for science in our universities. The Humanities postmodernists hold a worldview that is totally incompatible with the one upon which science is based, and they are slowly outflanking scientists in positions of power inside our universities. You will be aware of the ‘science wars’, if only from parodies such as Sokal’s and after, and the position has only worsened since; but academic scientists either just get on with their research or they give up, after challenging the Humanities worldview and receiving flack and realising the magnitude and intensity of the counterpush necessary. This is not going to end soon, and as things stand it is not going to end well. It is also largely a battle within the Left: between old Left (modernist university scientists) and new Left (the postmodernist humanities). In this battle the old Left would do well not to be scornful of non-Leftist allies who also deplore postmodernism, for the greater rift is between postmodernism and everything else.

      • Good points. There are other areas where the Old Left and non-left are closer than Old Left and New Left (dare I say “the Woke”?), though sometimes with the same goals for different reasons.

  3. While I’m at it, a puzzle. Suppose you have a plot similar to the one above, i.e. points (with (x,y) coordinates) in a plane. How can one find the smallest ellipse which encloses a given fraction of the points?

  4. Scientific American should know better. I thought there must be some misunderstanding, but upon reading the article it appears that they meant exactly what they said.

    Here is an article from the research literature that reaches this conclusion based on different data:

    For those who can’t or don’t care to access the full article, here’s the relevant figure:

    It looks to me like their data set actually does have some variation — that is, the parabola is a better fit than a constant would be — and the article is reasonably forthright in saying that the evidence is weak. My chi-by-eye says that, based on the points plotted in this article, the decline at the left is real, but the rise at the right may not be.

    • telescoper Says:

      To be really convincing they need to take the Fourier Transform.

      • I think I’d do a Bayesian evidence calculation. Compare the model y = constant with linear, quadratic, etc., and ask whether the Bayesian evidence favors anything but the simplest model (i.e., whether the improved goodness-of-fit is enough to justify the additional complexity of the model).

    • Supernaut Says:

      “Scientific American should know better.”
      SA has seen a sad decline in quality over the past several years.

    • Or they could have done a bootstrap or some other cross-validation or some other cursory overfitting checks. They could have compares the linear model with theirs based on some information criterion to see if the additional parameter is justified by the variance it explains. There are certainly more I haven’t thought of.

      I don’t dare read the article to find out if they did any of that.

  5. Reblogged this on Symptoms Of The Universe and commented:
    The exchange with Alessandro Strumia rumbles and lumbers on in the comments sections of previous posts, while over at his “In The Dark” blog, Peter Coles highlights that credulous over-interpretation of gender gap data is not the sole preserve of aggrieved and ideologically-biased particle physicists…

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: