🍾 Thought bubbles: Germany, effect sizes, shared political issues
I’m in Heidelberg, enjoying a week visiting my friend Christoph Vanberg on the banks of the Neckar, staying in an Airbnb with the book collection of a true Bildungsburger: literature in four or five languages, philosophy from Marx to Derrida, you name it. The dog is reading Hegel, in the original of course. I am reading Max und Moritz.
I like Germany partly the same way I like Norwich: it’s far enough from the centre of the action to be interesting, Norwich by sheer distance from London, Germany because of the language gap with the US. Hot concepts from The Discourse take a little while to make it over here; you feel that they are late to the party and don’t always get the joke. But Germany is big enough to be a cultural exporter in its own right. Some favourite German things:
The film Toni Erdmann. Nacktparty!
“Es ist mittwoch, meine Kerle”: cartoons by SmallLebowsky on Reddit.
The Museuminsel in Berlin, probably the best collection of museums in the world.
Schwarzbier is the same colour as Guinness, but to my mind much tastier.
Rainald Grebe, a comic/singer. All in German, sorry:
Natural selection effect sizes
I’ve worked a bit more on the HRS data. A big question is how large the effect of natural selection is, see earlier. Here I’m just inflating effect sizes to compensate for the noise in the measured polygenic scores. Roughly speaking, if polygenic scores are “true score plus pure noise”, then the effects of the true score and the measured score on anything are in the same fixed proportion. So to get the effect of the true score on RLRS, you just scale up your estimated effect by the ratio between the effects of the measured score and the true score on their target phenotype. With luck, (a) heritability statistics tell you the effect of the true score on the target phenotype; and (b) you can check the effect of the measured score on the target phenotype within the survey data.
Coefficient on RLRS is how a given polygenic score will affect Relative Lifetime Reproductive Success — basically, the number of children a person has compared to everyone around them. The lines are 95% confidence intervals. “Chip-heritability” means we’re estimating the effect of the true polygenic score that is captured on a DNA array chip, which measures a large number of common genetic variants. “Twin-heritability” means we’re estimating the effect of all genetic variation whatsoever.
There are lots of things that can go wrong here, especially with twin-heritability, because variants not on the DNA chip may not behave like variants that are on the chip. But it’s the best one can do using this data. Overall, I think the results mostly show our ignorance. For most of these scores we can’t rule out tiny effects close to zero, but we also can’t rule out effect sizes of 0.1 or 0.2.
What effect sizes are large enough to matter? A simple division is between small, medium and large. Small we don’t need to care about; large means that natural selection is proceeding at such a rate or scale that we should panic. Medium means something like “at least be aware of the issue and think of how policies are affecting it”. I don’t see much evidence for large effects. A rough cutoff for medium effects might be 0.1 standard deviations. At that rate, 54% of the next generation is below the mean for the previous one. For most of these scores, we can’t rule that out, at least for twin heritability. But note that the scores for educational attainment, which are the most accurately estimated, are small in magnitude and we can be fairly sure we’re below that 0.1 cutoff.
I continue to think that the underlying issue here is “we are raising many of our children such that they are unlikely to be healthy or well-educated”. Conditional on that being true, it’s not obvious to me whether it is worse if the underlying reasons are genetic or environmental:
Back to the UK tomorrow. My main political impression from this trip was that issues across Europe are very synced. We are all worrying about house prices, energy prices, the green transition, and war. Britons, insular as ever, probably haven’t noticed how many of these issues are shared across the continent.
If you enjoyed this, you might like my book Wyclif’s Dust: Western Cultures from the Printing Press to the Present. It’s available from Amazon, and you can read more about it here.
You can also subscribe to this newsletter (it’s free):