This week's educational rant
is brought to you by the letter H2.
Heritability is a tricky concept, and to – for once – give science journalists their due, they seem to have tried hard to explain it in this story. But it’s a concept that really needs a careful and quantitative explanation, and sadly this is not the kind of numerate utopia in which news media present mathematical concepts that aren’t shit. Blogs can, though, so here goes.
To work out a value for heritability, you start with a set of people (or sheep, or rosebushes or whatever), and you look at the variation in weight (or wool yield or leaf length or whatever you’re measuring). Then you try to work out how much of that variation is due to genes. Because heritability is the variation due to genes divided by all the variation in the population, it’s a statistic that’s specific to the set of people you’re measuring, in the environment they’re in. Put the same people somewhere else and measure’em again in five years and you might get a different number. Measure a different set of people, and you might get a different number still.
As an example, in Britain as a whole, the heritability of hair length will be low, because there’s a lot of variability in how often and how short we cut our hair. Among a population consisting entirely of Sikhs, who tend not to cut their hair, heritability of hair length would be considerably higher because there’d be less variation overall.
Also, a high heritability does not mean that a characteristic is fixed or unchangeable, because it doesn’t tell us anything about what happens if the environment changes. I could keep a bunch of cacti in my window-box in which the heritability of plant height was 100%, but that wouldn’t mean they’d all stay the same size if I transplanted them to a greenhouse.
So you can’t just find out the heritability-of-obesity-full-stop. If you measure the heritability of obesity in ten-year-old twins in twenty-first century Oxfordshire, it may or may not be valid to apply your result to the rest of the children in Britain, or the rest of the ten-year-olds in Western Europe. It’s almost certainly not valid to use it to predict what’d happen to children in a hypothetical future with daily exercise sessions or a tax on corn syrup.
Now, taking off my Genetics Expert Hat and speaking only as a statistically literate pinko commie Islamofascist sympathiser, I suspect that if we curtailed traffic, encouraged walking and cycling, stopped subsidising intensively farmed meat, did something about the long hours, low wages culture that leaves people little leisure time to play sports or cook properly, kept junk food pushers away from children as much as possible, and found some way to stamp out the self-destructive hatred many people feel towards their own bodies… well, it probably wouldn’t make many fat people thin, because loads of fat people spend their entire lives trying to get thin with only occasional or intermittent success. But:
a. I could be wrong, because we simply don’t have the kind of data that would enable us to make that kind of prediction, and
b. we should do it anyway, because it would almost certainly make us – whether we’re fat or thin – healthier and happer.
Heritability is a tricky concept, and to – for once – give science journalists their due, they seem to have tried hard to explain it in this story. But it’s a concept that really needs a careful and quantitative explanation, and sadly this is not the kind of numerate utopia in which news media present mathematical concepts that aren’t shit. Blogs can, though, so here goes.
To work out a value for heritability, you start with a set of people (or sheep, or rosebushes or whatever), and you look at the variation in weight (or wool yield or leaf length or whatever you’re measuring). Then you try to work out how much of that variation is due to genes. Because heritability is the variation due to genes divided by all the variation in the population, it’s a statistic that’s specific to the set of people you’re measuring, in the environment they’re in. Put the same people somewhere else and measure’em again in five years and you might get a different number. Measure a different set of people, and you might get a different number still.
As an example, in Britain as a whole, the heritability of hair length will be low, because there’s a lot of variability in how often and how short we cut our hair. Among a population consisting entirely of Sikhs, who tend not to cut their hair, heritability of hair length would be considerably higher because there’d be less variation overall.
Also, a high heritability does not mean that a characteristic is fixed or unchangeable, because it doesn’t tell us anything about what happens if the environment changes. I could keep a bunch of cacti in my window-box in which the heritability of plant height was 100%, but that wouldn’t mean they’d all stay the same size if I transplanted them to a greenhouse.
So you can’t just find out the heritability-of-obesity-full-stop. If you measure the heritability of obesity in ten-year-old twins in twenty-first century Oxfordshire, it may or may not be valid to apply your result to the rest of the children in Britain, or the rest of the ten-year-olds in Western Europe. It’s almost certainly not valid to use it to predict what’d happen to children in a hypothetical future with daily exercise sessions or a tax on corn syrup.
Now, taking off my Genetics Expert Hat and speaking only as a statistically literate pinko commie Islamofascist sympathiser, I suspect that if we curtailed traffic, encouraged walking and cycling, stopped subsidising intensively farmed meat, did something about the long hours, low wages culture that leaves people little leisure time to play sports or cook properly, kept junk food pushers away from children as much as possible, and found some way to stamp out the self-destructive hatred many people feel towards their own bodies… well, it probably wouldn’t make many fat people thin, because loads of fat people spend their entire lives trying to get thin with only occasional or intermittent success. But:
a. I could be wrong, because we simply don’t have the kind of data that would enable us to make that kind of prediction, and
b. we should do it anyway, because it would almost certainly make us – whether we’re fat or thin – healthier and happer.
Labels: Boffinry, Cakes And Ale, PCIFS
posted by MissPrism at 9:41 AM
3 Comments:
Amen to that! When I read stories like that bill that was submitted in Mississippi that would outlaw obese people in restaurants (all in the name of "saving healthcare dollars", of course), and it seems clear that politicians want to take the cheap route of shaming and blaming obese people, rather than taking steps that will actually improve people's health. It seems sensible to me that making fresh foods more available and affordable and creating walking-friendly communities would go a long way toward improving the nation's health. It's just easier for politicians to say "stop eating" rather than doing anything productive.
On heritability, it's interesting the extent to which some people go to reduce all of human activity to the codes in our genes. I got a bit angry about it last week.
Did you read about the huge million-dollar project in the UK schools designed to reduce obesity by getting kids to eat one more piece of fruit or vegetable a day? Not just to increase health, but specifically to reduce obesity. I read about this in New Scientist and nearly started foaming at the mouth I was in such a tizzy. How can people be so stupid?
Post a Comment
<< Home