Crémieux

Against Attribution Studies > An “attribution study” is the name I’ve given to studies that purport to provide estimates of important real-world quantities through extrapolations from estimates in other, uninformative studies. There are a lot of these papers and they often come with conclusions like ‘Gas stoves responsible for 13% of American asthma’, ‘68,000 die each year due to health insurer denials’, ‘DOGE killed a million billion Africans with its program cuts’. > This style of study is not inherently bad. There are ways to do it right, but they require more effort than multiplying what are oftentimes pretty random numbers together. https://www.cremieux.xyz/p/against-attribution-studies

In the past decade, research suggests that testosterone levels have actually gone up[1]. [1]https://www.cremieux.xyz/p/why-are-testosterone-levels-rising

Difficult topic, too much for a HN comment. Some good starting points: https://www.cremieux.xyz/p/brief-notes-on-scientific-critique/comment/…

Similar thing, permanent anti-cavity bacteria: https://www.cremieux.xyz/p/the-rise-and-impending-fall-of-the The patents expired, so nobody can raise the hundreds of millions to do the phase II/III safety trials.

https://www.cremieux.xyz/p/from-pollution-to-solution I understand your concern but I don’t believe the impact is as severe as you think.

Like many things where what constitutes base data changes and hence graphs change pretty radically, this can be cleanly explained by diagnostic drift. Here's a detailed breakdown: https://www.cremieux.xyz/p/how-to-end-the-autism-epidemic

An even better post here on how the Flynn Effect and Reverse Flynn Effect was never real in the first place: https://www.cremieux.xyz/p/the-demise-of-the-flynn-effect

That finding is probably not reliable because of the way they do binning: https://www.cremieux.xyz/p/brief-data-post?open=false#%C2%A7nonlineari…

Link to the related blog post by the author: https://www.cremieux.xyz/p/food-deserts-are-not-real The paper referenced in the x-tweet is "Food Deserts and the Causes of Nutritional Inequality", which can be found eg here[1]. [1]: https://web.stanford.edu/~diamondr/FoodDeserts.pdf

And there's claim that lead is maybe overrated after all https://www.cremieux.xyz/p/who-gets-exposed-to-lead

Have you read this post about p-value suspiciousness as a means of testing for widespread non-replicable results? https://www.cremieux.xyz/p/ranking-fields-by-p-value-suspiciousness And that's not to speak of things like biotech which is full of photoshopped images of fluorescence and whatnot.

There’s sort of no point in discussing this article since the full text is paywalled. The Gizmodo summary seems to suggest that it’s mostly just a meta review plus some extrapolation: they averaged some studies to find the typical effect on mental health and then multiplied it by all the lead exposure in the country. I’ll suggest an interesting article to read instead. I don’t have the expertise to evaluate if it’s true but it is a good depiction of where p-hacking and correlation-causation mixups could lead. https://www.cremieux.xyz/p/who-gets-exposed-to-lead

> or rather the byproducts of the digestion of sugar by bacteria in your mouth) ...for now. There may be a "vaccine" for dental caries if this recently HN'd article is to be believed and isn't sponsored content: https://www.cremieux.xyz/p/the-rise-and-impending-fall-of-the Though https://en.wikipedia.org/wiki/Caries_vaccine suggests it may not be that simple.

> He's trying really hard to disprove Haidt's post by poking holes in many of the studies. Because this is how evidence based reasoning works. If the evidence that is supposed to support the hypothesis is fundamentally flawed, then our hypothesis doesn't actually have any support. The fact that we can apparently so easily find flaws in what is supposed to be empirical evidence should make us more cautious about drawing firm conclusions. In fact, low quality literature is something of a plague in many social sciences at the moment (e.g. the replication crisis in social psychology). This post presents a decent discussion of this sort of issue https://www.cremieux.xyz/p/beware-the-man-of-many-studies

> IQ only weakly correlates to the intelligence […] it tests a bunch of stuff completely unrelated to intelligence (e.g. cultural trivia). These are unfortunate myths. See the following on IQ testing, relationship of various tests to general intelligence, and the absence of cultural bias in testing: - Standardized exams measure intrinsic ability, not racial or socioeconomic privilege https://milkyeggs.com/society/standardized-exams-measure-intrinsic-abi… - Fair and Square: A Conclusion on IQ Test Bias https://menghu.substack.com/p/fair-and-square-a-conclusion-on-iq-test-… "Generally, the most commonly used IQ tests aren’t biased or only minimally biased as to be of no practical value." - Bias is Often Unpredictable https://www.cremieux.xyz/cp/140810182 "Bias does not simply mean that a measure gives different scores to different groups. This is one of the most common misunderstandings about “bias”. What it really means is that members of different groups obtain different scores conditional on the same underlying level of ability." - The Relationship Between the Scholastic Assessment Test and General Cognitive Ability https://sci-hub.st/https://journals.sagepub.com/doi/abs/10.1111/j.0956… "Furthermore, empirical evidence suggests a substantial relationship between the SAT and g. In a study of 339 undergraduates, Brodnick and Ree (1995) used covariance structure modeling to examine the relationship between…

The author discusses this in footnote 3: https://www.cremieux.xyz/p/ranking-fields-by-p-value-suspiciousness#fo… > Alternative explanation I: Nonsignificant results may not be reported at all and thus they won’t appear in the dataset. > This cannot explain the results. Remember that when results should be significant because you have adequate power to detect a real effect, only a small percentage of p-values will be in the 0.05 to 0.01 range. People should be reporting p-values that are much lower for typical significant effects, but they aren’t: they're reporting suspicious effects. Publication bias is usually a bias towards significance not a bias towards a p-value between 0.05 and 0.01. People adjust their marginal estimates of interest to get at least below 0.05 but they don’t take their extremely significant estimates and adjust them so they’re in the suspicious range.

Take the author's interpretations and normative suggestions with a grain of salt. Cremieux personally promotes scientific racism (like in an article called "Elites are genetically different" [1]), defends discredited data from eugenicist Richard Lynn [2][3], and co-wrote a paper called "Global Ancestry and Cognitive Ability" with Bryan Pesta [4], Emil Kirkegaard ("Eugenics or Western civilization dies" [5]), and John Fuerst ("all major Jewish organizations... favor the dissolution of their host nations through massive non-White immigration" [6]). Side note: In an article questioning why Zohran Mamdani of Indian descent born in Uganda would mark checkboxes for "Asian" and "Black or African American" among the limited options in a university application [7][8], The New York Times cited Cremieux without even hinting at his past writings, then edited the article to describe Cremieux as "an academic" who "writes often about I.Q. and race". As Marc Channick pointed out [9]: > This is basically like describing a Pizzagate promoter as someone "who writes often about child safety" [1] https://www.aporiamagazine.com/p/how-do-elite-groups-form [2] https://www.cremieux.xyz/p/national-iqs-are-valid [3] https://www.splcenter.org/resources/extremist-files/richard-lynn/ [4] https://www.amren.com/news/2022/10/racial-pseudoscience-on-the-faculty… [5] https://archive.is/URqek - archived version of https://twitter.com/KirkegaardEmil/status/1005539088546529280 [6]…