Links 3/16: Rulink Class

New research: toxoplasma’s effects on cats and animals are as creepy as ever, but it probably doesn’t affect human behavior.

New drug nilotinib looks very promising for Parkinson’s disease, may clean up proteins associated with death of dopamine-producing cells. Good news: drug is already approved for cancer and so can be used off-label. Bad news: drug costs $10,000/month.

Vox has a pretty good article on Silicon Valley Democrats. Historical point of comparison: Theodore Roosevelt.

The Atlantic highlights the 1996 Dole/Kemp Campaign Website and the 1996 Clinton/Gore Website. Fricking Clinton/Gore ’96 launched a popup that tried to install Norton Antivirus on my computer. That’s a little more nostalgia than I’m ready for right now.

Ben Goldacre et al’s crusade against outcome switching in clinical trials: “So far, they’ve checked 67 clinical trials. Of those, nine trials were perfect. But among the ones that weren’t, they found 301 pre-specified outcomes were never reported and 357 were silently added.” Related: once the regulatory agencies required that pharma companies pre-register their trials, the positive finding rate dropped from 57% to 8%.

Latest meta-analysis finds homeopathy is effective for 0 out of 68 illnesses. Encouraging after some previous less rigorous studies got some unfortunate false positives in this area.

Wikipedia’s Special: Nearby gives you all the Wikipedia pages about places close to you. I got my local district library, but maybe people who live in more interesting places will get more interesting articles.

Asian cop shoots black victim in high-pressure situation. Black community protests that not imprisoning the cop would be racist against blacks. Asian community protests that imprisoning the cop would be racist against Asians. It’s almost as if turning every incident involving a minority into a morality play about racism can go wrong.

Contrary to previous results, a new study suggests that calorie labeling does make people lose weight, especially men.

Brookings: Declining fluidity in the labor market probably not due to changing population or increasing regulation, possibly due to changes in companies and decreased social trust.

Weird Sun Twitter is now a card game. Warning: card game might not actually be playable, relationship with Weird Sun Twitter unclear.

Nathan Robinson: Sanders would be uniquely good at campaigning against Trump, Hillary uniquely bad at it.

Popehat combines strong free speech advocacy with a strong insistence that private censorship is disanalogous to public censorship – something I sort of argued against here. Now blogger Ken White clarifies and gives a little more subtlety on his position: “I’m increasingly convinced by the argument that [Twitter] has decided to offer a product aimed at a specific political group…[but] I classify Twitter’s action as bad customer service and as private speech I don’t like because of my conservative views…at least I thought those were conservative views. I mean, how can you argue that a bakery shouldn’t have to make a gay marriage cake, but Twitter should have to offer a platform to someone they think (not unreasonably) is a total douche?”

Some good science/statistics blogging about a recent paper against the paleo diet.

Status 451: What Is Neoreaction? Definitely one of the more helpful introductions in this genre, by which I mean it doesn’t obsessively focus on being as controversial/offensive as possible to the exclusion of everything else. Note that the term is still banned in the comments section here, so discuss it over there if you have to.

One plank of Obamacare penalizes hospitals if patients get readmitted with the same disease too quickly after being discharged. There was a lot of concern that this would lead to hospitals fudging things or even going as far as refusing to readmit patients who need it. A study in NEJM finds that the program seems to be going well, that readmission rates are genuinely down, and that it isn’t a result of hospitals cooking the books.

Jeff Kaufman: buses are 67x safer than cars. They’re also underused, partly because they’re annoying, partly because of safety features. There is room to trade off bus safety for bus convenience, which would make people take more buses, which would actually make them safer in the long run. Therefore we should make buses more dangerous.

Futility Closet: 1/a long series of 9s with one 8 in it gives you a decimal representation of the Fibonacci sequence, for some reason.

Gambler’s fallacy in decision-making. Just as a gambler who’s had a long string of losses might be more likely to expect a win next time, so a judge who’s had a long run of innocent people will be more likely to find the next person guilty.

Change of heart: journalist who reported Minnesota county was the worst place in America to live has now decided to move there.

Reddit: what’s the next big thing in terms of trends that will shape our future? Suggestions include lab-grown meat, organic plastics, and CRISPR.

Eroom’s Law – a straight-line, Moore’s law style relationship showing that the average pharmaceutical company dollar buys fewer and fewer new drug discoveries over time. Reason unclear but possibly involving lower correlation between the models on which the drugs are tested and real human bodies.

Interesting weird post-modern papers and articles: World Toilet Day is an example of neocolonialist white supremacy, evidence-based medicine is an “outrageously exclusionary…example of microfascism at play in the contemporary scientific arena”, and geologists need a feminist glaciology framework to make sense of “the relationships among gender, science, and glaciers”.

Despite my concerns to the contrary, generic drug prices are going down.

Big reputable poll of Florida residents finds 10% of them believe Ted Cruz is the Zodiac killer. But remember Lizardman’s Constant. A good thing to keep in mind next time someone finds a poll that says 10% of Donald Trump supporters support drowning puppies or whatever.

How to communicate securely with people on LSD through messages that sober people would not be able to read, just in case for some reason you want to do that. Related: exceptionally weird short story/essay/something-or-other about consciousness.

Pacemakers work better than placebo pacemakers, but placebo pacemakers still work pretty well.

80000 Hours: a summary of the literature on whether money makes you happier. Short version: a little!

When nativists start building walls, migrants start building battering rams. I hope this escalates to moats and trebuchets.

An interesting and balanced piece on unemployment benefits. Finds that extending unemployment benefits does make people submit fewer job applications, but that very fact means decreased competition and greater ability for people who want to go back to work to do so! As a result, extending unemployment benefits doesn’t increase unemployment much.

A lot of people here talk about the Griggs vs. Duke ruling that bans IQ tests in a job interview, but for some reason police can still get away with only accepting medium-IQ people as cops. Bonus: court case is a high-IQ guy angry at being rejected for the force; court tells him to take a hike.

Scientific American: John Horgan interviews Eliezer Yudkowsky. I thought it was a really well-done interview, great answers from Eliezer, and really funny comment from Eliezer’s wife Brienne.

doubleblinded.com is a sort of supplement company that will send you both real and placebo supplements and everything you need to perform a randomized controlled trial on yourself to see if the supplements really help you. Sure, you can probably do it cheaper on your own if you really try, but maybe having someone else take care of the trivial inconveniences will encourage this sort of thing.

Scientists have identified over 20% of the genes involved in autism. I didn’t realize we were that far along with understanding any kind of massively polygenic trait like that.

Lithium: still the best treatment for bipolar disorder.

Study: given identical patient descriptions, therapists were twice as likely to diagnose boys as girls with ADHD. Obvious relevance for all those claims like “Men/women are X times more likely than women/men to have such and such a psych disorder”.

A lot of people ask – is eliminating tropical diseases just band-aid charity? Won’t it just mean more people survive a little longer to be starving and diseased and need help later? The answer has always been that eliminating diseases improves people’s health, employment, education, and possibly intelligence, with lots of positive effects down the road. Here’s a good example: huge economic gains and human capital increases from America eliminating typhoid.

During the Holocaust, Protestants were more likely to rescue Jews in majority-Catholic areas, and Catholics more likely to rescue Jews in majority-Protestant areas. Maybe being a minority makes you more sympathetic to other minorities or less willing to go along with the government in general?

This month in the media: “caucus moderator in Nevada requests neutral translator” becomes “caucus moderator shouts ‘ENGLISH ONLY’ at Hispanics” becomes “Sanders supporters shout ‘ENGLISH ONLY!’ at Hispanics” becomes “Sanders himself attends caucus in Nevada to shout ‘ENGLISH ONLY!’ at Hispanics”. H/t @freddiedeboer, who did good work publicizing this as part of his “media is shilling for Hillary” special interest.

Having a disruptive student in your class decreases your adult earnings by 3%. One possible reason for private school advantage is that they can reject these students or keep them in their own special classes/groups apart from the kids who actually want to learn without getting yelled at. Taken at face value, this is a pretty strong testimonial to the power of education – apparently education is so important that even one variety of disruption to it can seriously impact your adult earnings.

But related: “Cross‐national data show no association between increases in human capital attributable to the rising educational attainment of the labor force and the rate of growth of output per worker. This implies that the association of educational capital growth with conventional measures of total factor production is large, strongly statistically significant, and negative…educational quality could have been so low that years of schooling created no human capital.”

Related-ish: at least in Sweden, starting school before age seven is not helpful and in fact is likely harmful. 2016 presidential candidates react by vowing to triple the budget for Head Start.

When Medicaid stopped covering Planned Parenthood, relevant pregnancies increased 27%.

Google Deep Dream (you know, the AI image filter that creates weird dog-shoggoth mixes out of everything) can be applied to videos now. Unfortunately, they were so preoccupied with whether or not they could that they didn’t stop to think if they should.

There’s been some discussion on the subreddit about a 2006 Sampson et al study finding that neighborhood effects explain a lot of crime disparities. But a recent Sariaslan et al study finds that neighborhood effects on crime disappear once you control for genetics. And Jaap Nieuwenhuis finds the neighborhood effects literature to be riddled with publication bias and questions whether the effect exists at all. What I want to know: are neighborhood effects by definition shared environmental effects? Does that mean the whole behavioral genetics literature is telling us they’re not real?

Widely-read econblogger converts to Christianity and goes full creationist. Interesting look at how a self-described rational economist can end up believing some pretty unusual things.

Some fierce infighting in psychology as a Harvard/UVa team including Daniel Gilbert and Gary King denounce the OpenScience project and the replication crisis it highlighted as bogus (paper, popular article). They have two main arguments: first, the “replications” were so different from the original studies that different results are unsurprising; second, that because of the way statistical power and confidence intervals work, OpenScience finding only 40% of studies replicating is consistent with 80-90% of the studies being correct, and in fact another replication attempt that found 85% replication rate would have said only 40% of its studies replicated if they had used the same (incorrect) statistical methods as OpenScience. But the pushback from psychologists and statisticians defending the existence of a replication crisis has been intense and highly convincing. Here’s a 45-author paper published in Science saying that “Gilbert’s very optimistic assessment is limited by statistical misconceptions and by causal inferences from selectively interpreted, correlational data” – but as usual, all the interesting stuff is on random blogs. Brian Nosek on RetractionWatch explains how Gilbert at al seriously exaggerated some of the differences between original studies and replications to the point of absurdity; The 20% Statistician says that “the statistical conclusions in Gilbert et al (2016) are completely invalid”, and The Hardest Science finds that Gilbert’s example of the the 85% replication rate dropping to 40% because of poor methods involves completely inappropriate cherry-picking of metrics. I admit my bias here but AFAICT the Gilbert paper is looking pretty questionable and the replication crisis seems as real as ever.

Very much related: a new very large study of ego depletion finds the effect does not exist. This is a pretty big deal: since its inception, almost a hundred studies have found evidence of ego depletion, and it’s become an entire subfield of psychology with people investigating all the different factors that make it stronger and weaker. If the whole thing just doesn’t exist and the entire literature about it is a mirage, that’s really damning. A Slate article on the issue very kindly links my review of Baumeister’s book where I raised some of these concerns last year. Neuroscientist and ego depletion expert Michael Inzlicht writes an intense soul-searching essay: “I have spent nearly a decade working on the concept of ego depletion…I’m in a dark place. I feel like the ground is moving from underneath me and I no longer know what is real and what is not”. He adds that he suspects his other research area of stereotype threat may be heading in the same direction, and says that “During my dark moments, I feel like social psychology needs a redo, a fresh start.” Some more discussion on Beeminder forums.

Also related: peak-end effect fails to replicate.

Trump, Trump, Trump, Trump, Trump, Trump, Trump, Jeb, Trump, Trump.

If these aren’t enough links for you, wettrew on the SSC subreddit is collecting his own links roundups.