Noah Haber

@NoahHaber

Econ, epi, stats, meta, causal inference mutant scientist / epistemic humility fairy godmother, chaos muppet. exploring @noahhaber@mastodon.online

United Statesmetacausal.comJoined December 2011

1,304 Following

9,527 Followers

Noah Haber’s Tweets

Pinned Tweet

Noah Haber

@NoahHaber

Aug 30, 2021

Causal language study is now up on medRxiv! medrxiv.org/content/10.110 Ever wondered what words are commonly used to link exposures and outcomes in health/med/epi studies? How strongly language implies causality? How strongly studies hint at causality in other ways? READ ON!

287

791

Replying to

Every complex question has a simple, straightforward, easy to understand wrong answer

Noah Haber

@NoahHaber

11h

... or put another (and much pithier) way:

Quote Tweet

(wannabe) Ƀreaker of (the Bad) Loops

@generativist

14h

"The design principle that attention is scarce and must be preserved is very different from a principle of 'the more information the better.'" Herbert Simon [sent daily as a rhythmic reminder]

Re-surfacing this old thread as a reminder to myself: more information provided DOES NOT mean more information received. Also, as a reminder that I need to buy a new rope.

Quote Tweet

Noah Haber

@NoahHaber

Jan 9, 2020

I want to talk about the obscure statistic that changed my thinking about stats forever: The UIAA falls rating for testing climbing ropes A thread!

Show this thread

Controlling for height, Taylor Swift is as tall as Bruno Mars

257

Noah Haber

@NoahHaber

Dec 20

Shameless plug on a widget-filled piece I wrote regarding why ignoring uncertainty leads to poor decisions and some ways of dealing with it, in the context of GiveWell's cost-effectivess model-based selection process. (also it won a thing)

metacausal.com

GiveWell’s Uncertainty Problem

Noah A. Haber noahhaber@gmail.comORCiD: 0000-0002-5672-1769 Links:Code and data repositoryModified model sheet for PSA demoWeb app: Two programs with uncertaintyWeb app: Conceptual demonstrati…

Noah Haber

@NoahHaber

Dec 15

i hereby renounce my prize, this is the true winning post

Quote Tweet

Greg Sasso

@gregsasso

Dec 15

Replying to @NoahHaber

Have you considered buying a castle?

that's a problem for tomorrow noah. today noah is just gonna chill.

as promised, the vast majority prize $ is going to get donated w/ feedback from the many folks who helped with this. very, very curious to read the other entries and responses. even more curious what GW and others do differently with this moving forward.

o hey, it me

Quote Tweet

GiveWell

@GiveWell

Dec 15

One first-place winner is “GiveWell’s Uncertainty Problem” by Noah Haber. The author argues that without properly accounting for uncertainty, GiveWell is likely to allocate funding suboptimally, and proposes methods for addressing uncertainty. (2/4) bit.ly/3V18oyO

Show this thread

149

Sample size calculations become a lot easier when you realize "Hey fuck it, I can just simulate this"

289

I am begging you please don't make me become an RCTs-only guy

Clear case for research co-design with consumers. Why would I *want* to "cure" my dislike of coriander?

Quote Tweet

Noah Haber

@NoahHaber

Dec 12

my effective altruism pledge is to donate 10% of my income to find a cure for people who think cilantro tastes like soap

Show this thread

Noah Haber

@NoahHaber

Dec 13

ngl this shitpost went all eugenicsy much faster than I would have expected

I accept the ambiguity in the sentence wrt whether we are being cured of the condition or the people for the good of future humanity.

620

infinite long run utility if you think about it

my effective altruism pledge is to donate 10% of my income to find a cure for people who think cilantro tastes like soap

First big project in my shop finally starting to take shape (albeit upside down).....

What are some good citations on rollout of medical innovations in US versus peer nations? A lot of treatments are available first in US and initial uptake is quicker but highly inequitable. Other countries often eventually surpass US in population coverage.

Heads up that

is doing our monthly #PreprintCommentClub this afternoon where Arcadians set aside time to comment publicly on preprints. We’re reading preprints anyway so want to help authors improve their work and readers gain context.

Periodic reminders: 1) People who are very successful in bad systems are not likely to do much of anything to change that system. 2) You are exceedingly unlikely to become the exception to that rule, should you make it in that system too.

Noah Haber

@NoahHaber

Dec 8

wenge is beautiful, but love is pain.

so. many. splinters.

ITS LIKE RIGHT THERE. Get on it psychologists.

How on earth has no one written an article headlined "Fiedler on the Roof"

did some poking around restricting it to certain fields, didn't see any notable differences in trends (just levels). 30+ years and thousands upon thousands of risk factor studies, and no one can tell me what the hell they even are.

source: quick pubmed search, journals selected from the 18 journals in our causal language project thing, data/full terms here:

docs.google.com

risk factor search term

Sheet1 Year,count with "risk factor",count, all,All 18 journals,Year,% of abstracts/titles 2022,133,4710,2.82%,1972,0% 2021,118,5527,2.13%,1973,0% 2020,137,5305,2.58%,1974,0% 2019,120,5052,2.38%,19...

well this is pretty interesting...

3" machinists square for reference

omg it gets better

has AI gone too far

i can't stop looking at this and now you have to too

He criticized another field without knowing much about the method so badly we almost had to make him an honorary economist.

426

Some context: an EIC of a major epi journal recently claimed that paying reviewers would not improve review quality or quantity substantially and/or that it would not be practical, and that ECRs should just review more to solve the problem. * not naming, but is verifiable

to the "but we can't afford to pay reviewers" crowd: twitter.com/Andrew___Baker

This Tweet is unavailable.

Chat consensus seems to be that the "causal revolution" has had negligible impact on epidemiology as a field as a whole, where the impact has (so far) been confined to a small insider set and maybe a few subfields.

Also not necessarily a critique; tangible institutional change occurring in time periods less than generational is relatively rare in academia. I can only name 2(ish) socially-oriented fields that have had genuinely major tangible change measurable in single digit years.

Genuine question btw; I am genuinely interested in institutional change. Very likely that I am missing things, that the impacts are tanglible in some sub-fields, that they are happening but not yet observable, etc who knows.

What am I missing? Where else should I be looking?

Show this thread