Men have an advantage in strength that goes beyond being more muscular.

We know this because they tend to be stronger even when members of both sexes are matched for muscle mass. I separated off these results because of the difference in scale. The finding holds up with these measures too.

Interesting paper.

Because Taiwan randomly assigns students to classrooms (max size = 35 kids), these researchers were able to estimate that adding one Formosan child to a classroom reduced the test scores of the other students by the equivalent of at least 0.18 IQ points.

https://twitter.com/nberpubs/status/1790457271798902997

And to get ahead of it: there was no evidence that indigenous students helped other indigenous students, as the coefficient on the interaction was anything but significant (p = 0.82).

Now look at these peer-level mediators:

The relationship between cognitive ability and the rate at which people experience cognitive decline is null.

That means that people with low and high levels of cognitive ability decline at similar paces, just from different levels.

Look at the meta-analytic relationship: A similar finding is that education doesn't moderate the pace of cognitive decline.

Here are meta-analytic results for declarative memory as an example. The relationship with outliers clocks in at -0.0003, and without them, at 0.0005, both null.

One way to see if social media is polarizing is to take it away.

A new large, preregistered study of social media deactivation before the 2020 election looked at exactly that🧵

The design involves randomizing thousands of users to deactivating their accounts: Because social media can be watched, it's easy to see compliance

We can also ask if people substitute into other social media, which might have different echo chambers

Both deactivation groups substituted into using other social media, covering 26% (FB) and 57% (IG) of the gap.

As generations are declining in size, so too are the sizes of elementary, middle, and high school classes.

🧵 Universities have also been hit, and that's despite the percentages of students enrolled in a given age increasing.

Some economists tried to reproduce the results of 67 economics papers and they pretty much couldn't do it:

Even with help from authors, only half of papers ended up being reproducible, and this was still a problem at journals with required reporting of code and data. An earlier paper looked at the evidence for reproducibility in international development impact evaluations.

Similar story there: too much data was simply unavailable and some was discrepant.

Peer reviewed publications overstate how well anti-depressants work because the published literature omits lots of conflicting results🧵

When the FDA does their reviews, they notice lots of unpublished studies that tend to show the drugs are less effective. The sizes of the effects observed by the FDA and in journals differ, in some cases substantially, and in others only slightly.

But overall, the effect sizes associated with antidepressants are 32% smaller in the FDA's reviews than they are in peer reviewed publications.

Because TikTok is controlled by the Chinese Communist Party, they suppress topics that are politically sensitive in China.

🧵 When it comes to Uyghurs, there's Uyghur content, but on TikTok, it tends to be much more often about their culture than about the abuses they're suffering under the Chinese state.

Does diversity make teams work better?

Apparently not!

A new, comprehensive preregistered meta-analysis found that, whether the diversity was demographic, cognitive, or occupational, its relationship with performance was near-zero. These authors were very thorough

Just take a look at the meta-analytic estimates. These are in terms of correlations, and they are corrected for attenuation

These effect sizes are significant due to the large number of studies, but they are very low, even after blowing them up

Mathematics education is a great way to see genetic stratification happening.

Individuals with higher educational attainment polygenic scores tend to do higher-level mathematics courses and persist with mathematics education for longer. There also seems to be some moderation of this effect by school quality.

For instance, comparing high-quality and low-quality schools, persistence (going beyond geometry) is about the same for high PGS (low PGS) students, but there are differences at lower (higher) PGS.

What sorts of policies stop crime?

Policies that reduce material hardship can claim to have some effects on property crime, but the evidence that policies can reliably make a dent in violent crime is much weaker. And this makes sense: violent crimes are typically crimes of passion, where someone goes into a rage in a moment, with no real, coherent motivations.

What's more convincing?

p = 0.04 in a sample of 10 or p = 0.04 in a sample of 1,000,000?

🧵 Pick an answer, then go to the next post.

What's likely? What's unlikely?

If chances are slight, what are the odds?🧵

If you ask people to assign numbers to these terms, here's what you get: What's a lot? If you have hundreds of something, is that nearer to 100 or 900?

If you ask people about those sorts of quantities, you can also get a glimpse at what the typical man means when he uses them:

A new study of the relationship between intelligence and politics just came out.

The relationship everyone knows holds up: more intelligent people tended to have more left-wing views!

Let's explore🧵 The first thing to note is that this is from an adoption study. That provides a lot of analytic leverage.

We can readily observe that this relationship is not due to indirect parental genetic effects, nor is it due to the family environment.

Expressing the effects of different medications after converting them to correlations might help people to overcome their reflexive disdain for small but reliable correlations, since "small" effects are frequently extremely meaningful.

Here are some examples: The need to place effects in practical terms is particularly pressing when a very large effect in the real world takes on what seems to be a small in terms of common effect size criteria.

It's important to remember, those criteria are arbitrary.

A recent study looked at Big Five personality differences between people suffering from 'Long COVID' and matched controls who had never been infected.

Long COVID sufferers were less agreeable, conscientious, extraverted, and open, but more neurotic. This isn't the only study to have investigated the relationship between personality and response to COVID.

Kohút, Kohútová and Halama (2021) found that more neurotic Slovaks did more food stockpiling, spent more time searching for info, and responded more negatively.

Paleo artists often "shrink wrap" fossilized animal depictions

The T-Rex, Utahraptor, Triceratops—popular depictions of each of these animals shows skin so close to bone that it might be unrealistic

So let's shrink-wrap existing animals🧵

Can you guess what this is? Preemptive note: All of this artwork is from C.M. Koseman, whose book (which is way more extensive than this thread) and some other material is linked at the end of the thread.

The last animal was a rhinoceros—the thing with the cooling heat sail!

Any idea what this animal is?

The American nuclear industry illustrates negative learning: the costs of plants have increased over time.

But this is not nuclear's fault. Almost everywhere else, the learning rate is positive: costs decline as the industry gains experience building!

🧵

Consider France: The U.S. has really only been experiencing cost overruns since the Three Mile Island incident, and the reason has to do with the industry becoming overregulated as a result of the public outcry that ensued.

Globally, almost 1% of people have been infected by a virus that can cause a form of leukemia, the human T-cell leukemia virus type 1, HTLV-1.

Among Australian Aboriginals, this disease is sexually transmitted and, as a result, almost half of those aged 45+ are infected by it. While it seems like every fact about Australian Aboriginals is some form of "Would you like to hear a horror story?", I still think this one is up there among the worst Aboriginal facts.

This group has the highest prevalence of HTLV-1 in the world, but they're not alone.

Among trials I'd like to see replicated is this one of the antibody bimagrumab.

In a sample of diabetics, it led to semaglutide-like weight loss with zero lean mass loss by end of study. In fact, lean mass went up, consistent with bimagrumab mimicking myostatin deficiency. I want a replication because

- Smallish sample, so we're far from sure
- We need samples of non-diabetics

1/2

Have you ever wondered why advertisements heavily feature Black people when they're only 12-14% of the U.S. population?

A new paper might have an explanation: Blacks have a strong preference for seeing other Black people in media, whereas Whites have no racial preferences. These results are based on meta-analyses of 57 and 76 pre-2000 effect sizes for Blacks and Whites and 112 and 87 post-2000 effect sizes.

If you look at these effects over time, it becomes clear that Whites' initial, slight racial preferences have declined and maybe reversed.