Wednesday, 10 July 2013

Science, girls, statistics - what could go wrong?

The use of statistics by the media is something that constantly drives me round the bend. (At least, it does 90% of the time.) Now the BBC has wound me up by combining science, gender issues and, yes, statistics.

To be fair these are not blatant errors, but rather that hoary old standard, not being scrupulous about separating correlation and causality. As we saw with the infamous high heels and schizophrenia study, even academics can be prone to this, but the media does it every day. One very common example is where they tell us on the news that the stock market went up or down as a result of some event. Rubbish. In most circumstances the stock market is far too chaotic a system to attribute a change to an event that happened around the same time. It's guesswork and worthless.

Here, the misuse is slightly more subtle. 'Girls who take certain skills-based science and technology qualifications outperform boys in the UK, suggest figures' says the relatively mild headline. But is this really what the figures say, and if so what should we deduce?

According to exam publisher Pearson, girls who take BTECs in science and technology are more likely than boys to get top grades. Now here's a key sentence. According to the BBC 'Despite this success, girls are vastly outnumbered by boys on these courses.' The implication here is that this is just the tip of the iceberg, and with many more girls we would have lots of better grades. The suggested correlation is of gender with good grades. However it could equally well be that this is self-selection, a regular plague on the houses of those attempting to interpret statistics. If there are large numbers of boys on the courses, many of them could be there because 'that's what boys do' not because they have any talent for the subject. By contrast, if there are a small number of girls (in this case between 5 and 38% depending on topic), then they are likely at the very least to have greater than average enthusiasm, and quite possibly greater talent. If this is the case, all this is saying is that 'better than average female candidates do well compared with average male candidates.' Not quite such a strong story - in fact not a story at all.

The article then goes on to quote someone saying too few girls take STEM subjects. Now, I think this is true. We still have an artificial cultural bias about girls going in for science and it is wrong. However, what we mustn't do is to try to support the belief that this is wrong with data that doesn't contribute anything to the argument. By putting the 'girls are better at it' supposed statistic alongside the desire to have more girls in the subject implies that there is something inherent in the gender that makes girls better at it, so we want more of them. No, no, no. We want more because girls should have the same opportunities, because they shouldn't be put off science/tech because their peers think it's inappropriate. Not because a dubious interpretation of stats implies we could improve the quality of our STEM stock of students because girls are better at it. Without effective evidence this is just as sexist as saying girls shouldn't do science because it's too difficult for their little brains.

A girl and a science building. See, they can go together! *
One last example from the article. We have a quote from Helen Wollaston of Women into Science and Engineering saying the results prove "that girls can do science, IT and engineering." That's a silly thing to say. Firstly there is nothing to prove. Why would they not be able to? But also, as we've seen, all these results seem to show is that the most motivated girls are better than the average boys. There should be no need to use dubious statistics to 'prove' that girls can do STEM. I don't think anyone has doubted this since we stopped thinking (as they genuinely once did) that these subjects would overheat delicate female brains. What we need to prove is that far more girls can be interested in STEM and that we can change the culture so that it is cool for them to do so. That is a totally different issue - but it is the real one we face.

* In the interest of openness and scientific honesty, I ought to point out that the woman portrayed was a music student.

No comments:

Post a Comment