the address seems a bit strange, we are standing in Brøndby... if the address is wrong and anyone sees this, then please reply :)

Domain: (Applied) Bayesian Statistics

Link: Statistical Rethinking (free pdf), My Less Wrong Review, The 2017-2023 Lectures*

Author: Richard McElreath

Type: Book, YouTube lectures and less wrong post about Bayesian Statistics books in general

Why: Modern Bayes relies on HMC sampling, this book goes all in on this approach, this allowed you to focus on how to build the model and allows you to skip all math (except for the link function), by sacrificing a little bit of mathematical rigor this book covers more than all other popular books on the subject, to the point where you can stop 2/3 way trough and consider the last 1/3 "advanced optional topics". 

*The 2017 and 2019 were great, I have not watched the newer versions of the course, the older versions of the book uses ulam a pedagogical STAN wrapper powerful enough for most of the the exercises in the book, written by the author, I would advice serious students to do all the models in Stan, pymc or a wrapper that uses those.

Yes, and EA only takes a 70% cut, with a 10% discount per user tier, its a bit ambiguously written so I cant tell if it goes from 70% to 60% or to 63%

Why the down votes?, this guy showed epistemic humility and said when he got the Joke, I can understand not upvoting as it is not the most information dense engaging post, but why down vote?, down voting confuses me and I fear it may discourage other people from writing on LW.

Edit: this post had -12, so probably 1-2 super down voted or something, and then stopped.

• Bronze User, 10€/month, gain Super upvote ability
• Silver User, 20€/month, posts cannot be down voted
• Gold User, 30€/month, post can be promoted to front page
• Platinum User, 50€/month, all posts are automatically promoted to the front page and curriated.
• Diamond User, 100€/month, user now only see adds on long posts

Loot Box: 10% Chance for +100 upvotes, 5% Chance for curriated status of random post

Each user tier gives 1 loot box per month.

Unable to comply, building in progress.

I am glad that you guys fixed bugs and got stronger estimates.

I suspect you fitted a model using best practices, I don't think the methodology is my main critique, though I suspect there is insufficient shrinkage in your estimates (and most other published estimates for polygenic traits and diseases)

It's the extrapolations from the models I am skeptical of. There is a big difference between being able to predict within sample where by definition 95% of the data is between 70-130, and then assuming the model also correctly predict when you edit outside this range, for example your 85 upper bound IQ with 500 edits, if we did this to a baseline human with IQ 100, then his child would get an IQ of 185, which is so high that only 60 of the 8 billion people on planet earth is that smart if IQ was actually drawn from a unit normal with mean 100 and sigma 15, and if we got to 195 IQ by starting with a IQ 110 human, then he would have a 90% chance of being the smartest person alive, which I think is unlikely, and I find it unlikely because there could be interaction effects or a miss specified likelihood which makes a huge difference for the 2% of the data that is not between 70-130, but almost no difference for the other 98%, so you can not test what correctly likelihood is by conventional likelihood ratio testing, because you care about a region of the data that is unobserved.

The second point is the distinction between causal for the association observed in the data, and causal when intervening on the genome, I suspect more than half of the gene is only causal for the association. I also imagine there are a lot of genes that are indirectly causal for IQ such as making you an attentive parent thus lowering the probability your kid does not sleep in the room with a lot of mold, which would not make the super baby smarter, but it would make the subsequent generation smarter.

Thanks, I am looking forward to that. There is one thing I would like to have changed about my post, because it was written a bit "in haste," but since a lot of people have read it as it stands now, it also seems "unfair" to change the article, so I will make an amendment here, so you can take that into account in your rebuttal.

For General Audience: I stand by everything I say in the article, but at the time I did not appreciate the difference between shrinking within cutting frames (LD regions) and between them. I now understand that the spike and slab is only applied within each LD region, such that each region has a different level of shrinkage, I think there exists software that tries to shrink between them but FINEMAP does not do that as fare as I understand. I have not tried to understand the difference between all the different algorithms, but it seems like the ones that does shrink between cutting frames does it "very lightly"

Had I known that at the time of writing I would have changed Optional: Regression towards the null part 2. I think spike and slab is almost as good as using a fat-tailed distribution within each cutting frame (LD region), because I suspect the effect inflation primarily arises from correlations between mutations due to inheritance patterns and to a much smaller degree from fluctuations due to "measurement error/luck" with regards to the IQ outcome variable (except when two correlated variables have very close estimates). So if I were to rewrite that section, I would instead focus on the total lack of shrinking between cutting frames, rather than the slightly insufficient shrinkage within cutting frames.

For an intuitive reason for why I care:

• frequentest: the spike and slab estimator is unbiased for all of my effects across my 1000+ LD regions.
• Bayesian: bet you 5$ that the most positive effect is to big and the most negative effect is to small, the Bayesian might even be willing to bet that it is not even in the 95% posterior interval, because it's the most extreme from 1000+ regions[1]. 

Not For General Audience, read at your own peril

Pointing at a Technical approach: It is even harder to write "how to shrink now" since we are now doing one more level of hierarchical models. The easiest way would be to have an adaptive spike and slab prior that you imagine all the 1000-2000 LD slap and spike priors are drawn from, and use that as an extra level of shrinkage. That would probably work somewhat. But I still feel that would be insufficient for the reasons outlined in part 2, namely that it will shrink the biggest effects slightly too much, and everything else too little, and thus underestimate the effects of the few edits and overestimate the effects of many edits, but such a prior will still shrink everything compared to what you have now, so even if it does insufficient/uneven shrinkage, it's still a better estimate than no shrinkage between LD regions.

Implementation details of 3-level spike and slab models: It is however even harder to shrink those properly. A hint of a solution would be to ignore the fact that each of the spike and slab "top level adaptive priors" influence both the slab and spike of the 1000+ LD shrinkage priors, and thus only use the spike to regularize the spike and the slab to regularize the slab. It might be possible to estimate this "post hoc", if your software outputs a sufficient amount of summary statistics, but I am actually unsure.

Implementation details of 3-level Gelman model: If you for some magical reason wanted to implement the method proposed by Andrew Gelman, as a two-level hierarchical model, then I can say from experience that when you have no effects, the method sometimes fails[2], so you should set number of mixtures to 1 for all LD regions that "suck" (suck=any mixture with one or more sigma < 1). I actually suspect/know the math for doing this may be "easy", but I also suspect that most genetics software does fancy rule-of-thumb stuff based on the type of SNP, such as assuming that a stop codon is probably worse than a mutation in a non-coding region, and all that knowledge probably helps more with inferences than "not modeling tails correct" hurts.

• [1] I am not sure this bet is sound, because if the tails are fat, then we should shrink very little, so the 1:1000 vs 1:20 argument would utterly fail for a monogenic diseases, and the spike and slab stuff within cutting frames does some shrinkage.
• [2]If statisticians knew how to convolve a t-distribution it would not fail, because a t-distribution with nu=large number converges to a normal distribution, but because he approximates a t-like distribution as a mixture of normals, it sometimes fails when the effects are truly drawn from a normal, which will probably be the case for a few LD regions.

One of us is wrong or confused, and since you are the genetisist it is probably me, in which case I should not have guessed how it works from statistical intuition but read more, I did not because I wanted to write my post before people forgot yours.

I assumed the spike and slap were across all SNPs, it sounds like it is per LD region, which is why you have multiple spikes?, I also assumed the slab part would shrink the original effect size, which was what I was mainly interested in. You are welcome to pm me to get my discord name or phone number if a quick call could give me the information to not misrepresent what you are doing

My main critique is that I think there is insufficient shrinkage, so it's the shrinkage properties I am mostly interested in getting right :)

if I had to guess, then I would guess that 2/3 of the effects are none causal, and the other 1/3 are more or less fully causal, but that all of the effects sizes between 0.5-1 are exaggerated by a factor of 20-50% and the effects estimated below +0.5 IQ are exaggerated by much more.

But I think all of humanity is very confused about what IQ even is, especially outside the ranges of 70-130, so It's hard to say if it is the outcome variable (IQ) or the additive assumption breaks down first, I imagine we could get super human IQ, and that after 1 generation of editing, we could close a lot of the causal gap. I also imagine there are big large edits with large effects, such as making brain cells smaller, like in birds, but that would require a lot of edits to get to work.