Sign in

Zach Shahn

@zshahn.bsky.social
146 followers 333 following 72 posts

Causal inference, assistant professor at CUNY SPH. All likes are endorsements, but maybe by my 3 year old who stole my phone.

PostsRepliesMedia
Zach Shahn @zshahn.bsky.social · 07/10/2026
Think that intuition about proofs in particular was already latent in you before the polarizing path to this point?
110
Zach Shahn @zshahn.bsky.social · 29/09/2026
Anyone want to see a way to handle a more general class of treatment strategies under more plausible conditional parallel trends assumptions? This paper with Jamie Robins and others has been out for a few years but completely ignored by economists arxiv.org/pdf/2204.10291
arxiv.org
010
Zach Shahn @zshahn.bsky.social · 25/09/2026
I interpret Idiocracy to be this
011
Zach Shahn @zshahn.bsky.social · 12/06/2026
Wait really? Or am I being tricked again?
111
Zach Shahn @zshahn.bsky.social · 12/06/2026
When I was a kid my dad convinced me that the actual word is 'flutterby', which obviously makes more sense
120
Zach Shahn @zshahn.bsky.social · 02/06/2026
This is such a great idea. Did they each choose their own papers?
110
Zach Shahn @zshahn.bsky.social · 20/05/2026
He didn't write that one though!
000
Zach Shahn @zshahn.bsky.social · 12/05/2026
Slightly increase a band Boyz 3 Men
010
Zach Shahn @zshahn.bsky.social · 07/05/2026
This actually kind of works though
000
Reposted by Zach Shahn
Matt Hill @gethill.bsky.social · 06/05/2026
Prime with ads. Just how Kubrick intended.
32964391368
Zach Shahn @zshahn.bsky.social · 14/04/2026
The corollary is that the impact of the real economy on politics has decreased (not vanished). I dont think political scientists have consensus on either point. These are also causal questions that it would be difficult to properly address by modeling a time series.
000
Zach Shahn @zshahn.bsky.social · 14/04/2026
I don’t think stancil stated his argument clearly. It really has two parts. 1. The impact of the economy on politics is mostly via voters’ opinions about the economy and not their personal circumstances. 2. The impact of the actual economy on people’s opinions has decreased due to social media.
100
Zach Shahn @zshahn.bsky.social · 26/03/2026
Difference in differences *is* an observational research design. It doesn’t assume no unobserved confounding, but is still based on observational data and makes other unverifiable assumptions. “quasi-experimental” studies arent inherently more plausible except to economists
011
Zach Shahn @zshahn.bsky.social · 18/03/2026
This is what drives the newly rich main character of Remainder insane with Naz playing the role of chatbot. Also AI psychosis is probably similar to billionaire psychosis
021
Zach Shahn @zshahn.bsky.social · 20/02/2026
Maybe a sub question: is art created by an llm inherently different than the same art created by a human and if so how?
010
Zach Shahn @zshahn.bsky.social · 20/02/2026
Deterministic arrows representing a function of covariates at the unit level (however you got that function) are fine. But the function itself as a random object is not at the same level as the units. If you want to include it in a dag, you need a dag where samples are the units and you have n of 1
000
Zach Shahn @zshahn.bsky.social · 20/02/2026
I think dags represent population level distributions and am having a hard time seeing how it’s coherent to include sample statistics at all… (maybe the letter addressed this, couldn’t access full article on phone)
100
Zach Shahn @zshahn.bsky.social · 22/01/2026
I know you intentionally didn't name them, but any chance you'd be willing to? Throwing them new readers like me might outweigh the mild public shaming?
010
Reposted by Zach Shahn
sadvil 😎 @sadvil.bsky.social · 16/01/2026
protestors have painted what appears to be a tunnel on the side of the mountain, causing ICE agents to run into the mountain at full speed. disgusting.
164119472313
Zach Shahn @zshahn.bsky.social · 08/01/2026
My internal monologue
050
Zach Shahn @zshahn.bsky.social · 09/12/2025
A more borderline one is track/swim meets. You can break the world record in a qualifying heat and then get a middling time/score in the final heat and get no medal, or do the reverse and get gold
010
Zach Shahn @zshahn.bsky.social · 09/12/2025
More generally, in tennis you want to spread out your total points won so that you win the most games and sets. You can easily win more points and lose a tennis match if they're not properly grouped
050
Zach Shahn @zshahn.bsky.social · 03/12/2025
You should read Masters of Atlantis: en.wikipedia.org/wiki/Masters...
en.wikipedia.org
Masters of Atlantis - Wikipedia
010
Zach Shahn @zshahn.bsky.social · 18/11/2025
Waiting for the regression discontinuity
010
Reposted by Zach Shahn
Marco Piccininni @mpiccininni3.bsky.social · 11/11/2025
Thank you for the invitation and for the great discussion!
092
Zach Shahn @zshahn.bsky.social · 03/11/2025
We (Audrey Renson and @pausalz.bsky.social) Just updated this paper on interference in time-varying DiD settings from a while ago: arxiv.org/abs/2405.11781. Still see papers coming out regularly about problems that I'm pretty sure this solves...
arxiv.org
Structural Nested Mean Models Under Parallel Trends with Interference
Despite the common occurrence of interference in Difference-in-Differences (DiD) applications, standard DiD methods rely on an assumption that interference is absent, and comparatively little work has...
011
Zach Shahn @zshahn.bsky.social · 29/10/2025
Is there any movement to get them to stop showing players’ postseason stats on tv? I wanna know how good they are, not how lucky they were the past few games
000
Zach Shahn @zshahn.bsky.social · 26/10/2025
And I now see how this doesn’t depend on intervention. Eg conditioning on being on earth induces associations between variables related to falling explained mathematically by Newton. So I think I fully understand now, thanks for bearing with the stream of consciousness if you made it this far!
010
Zach Shahn @zshahn.bsky.social · 26/10/2025
(Didn’t Google Boyle’s law, hope I got it right)
100
Zach Shahn @zshahn.bsky.social · 26/10/2025
I guess it doesn’t need to be as magical as fate. You could imagine presetting a machine to maintain gas pressure at some level however it needs to. Then temperature and volume become counterfactually related. And P=TV is a mathematical non-mechanistic explanation like you discuss in the paper
100
Zach Shahn @zshahn.bsky.social · 26/10/2025
Ok, but now I’m back to my original understanding that the distinguishing feature of pre-selection is that it occurs via intervention and induces association via “fate” as in your fairy godmother example
100
Zach Shahn @zshahn.bsky.social · 25/10/2025
Ok, but C is in the past of E and D, the exposure and outcome of interest? I think I'm being dense, so I won't ask you to explain again!
110
Zach Shahn @zshahn.bsky.social · 25/10/2025
In M-bias, you’re conditioning on C, which is something pre-baseline like an eligibility criterion for a study. But it doesn’t arise from actually intervening to set C. Is that what makes predecessor bias different?
100
Zach Shahn @zshahn.bsky.social · 25/10/2025
Ha, I don’t think anything’s wrong, but I think M-bias from conditioning on pretreatment variables is one common case of what you’re describing
100
Zach Shahn @zshahn.bsky.social · 24/10/2025
Yeah I was asking about your pinned post paper, read it with great interest
110
Zach Shahn @zshahn.bsky.social · 24/10/2025
Does predecessor bias have to come from a 'fated collider' that seems to maybe be a uniquely quantum thing? Or do you use the term to refer to any setting where conditioning on a pre-intervention variable induces an association? We call the latter case 'M-bias' journals.lww.com/epidem/abstr....
journals.lww.com
110
Zach Shahn @zshahn.bsky.social · 24/10/2025
Causal people are usually interested in causal effects and would want to adjust away what we would call that spurious association. So we would just call it confounding adjustment. Maybe more descriptive people like demographers have a word for adjusting for confounding when you don’t want to
130
Zach Shahn @zshahn.bsky.social · 16/10/2025
Forgot to say, 'we' is me and my phd advisor David Madigan, who took a break from being a provost to do research again. Nice reunion!
000
Zach Shahn @zshahn.bsky.social · 16/10/2025
Next, we ask 'when will the key assumptions be satisfied'? We argue that under a causal pie model the answer is 'basically never'! But we think violations should be small in practice and provide sensitivity analysis.
100
Zach Shahn @zshahn.bsky.social · 16/10/2025
Then we develop Neyman orthogonal estimators for when S is only independent of Y(1) given Y(0) and covariates.
100
Zach Shahn @zshahn.bsky.social · 16/10/2025
First, we simply point out that you don't really need multiple studies. You just need one study and a baseline covariate S that you can use to make your own 'substudies' that satisfy their assumptions. This gives you more control over whether the assumptions are approximately satisfied.
100
Zach Shahn @zshahn.bsky.social · 16/10/2025
I think this paper we just put out is kind of interesting (arxiv.org/abs/2509.20506)! Wu and Mao (arxiv.org/abs/2504.20470) cleverly showed that if you have multiple studies and 'study' is associated with Y(0) but not Y(1) given Y(0), you can identify the joint distribution of Y(0) and Y(1).
arxiv.org
Identification and Estimation of Joint Potential Outcome Distributions from a Single Study
Most causal inference methods focus on estimating marginal average treatment effects, but many important causal estimands depend on the joint distribution of potential outcomes, including the probabil...
110
Zach Shahn @zshahn.bsky.social · 05/09/2025
Dare I say that this post is ironically on the verge of starting a lively debate?
010
Zach Shahn @zshahn.bsky.social · 07/08/2025
Ha, yeah, that was "unconstructive". Guess you can usually tell who wants feedback and who's celebrating from the wording of their post. Definitely shouldn't rain on people's parade when they're clearly celebrating
110
Zach Shahn @zshahn.bsky.social · 07/08/2025
I'm sure this a subtweet of something that I wouldn't defend, but I do post papers here in large part to hopefully get feedback...
100
Zach Shahn @zshahn.bsky.social · 06/08/2025
Basic idea is: you want to know the effect of some exposure on longevity (via MR). You don't know how long the people in your sample will live. But you know how long their parents lived. So you use sample's genes as IVs, sample's exposure, parents' lifespan as outcomes. When/why would this work?
000
Zach Shahn @zshahn.bsky.social · 06/08/2025
Does this paper (pubmed.ncbi.nlm.nih.gov/34669630/) address the issues raised in that paper?
pubmed.ncbi.nlm.nih.gov
Constructed Measures and Causal Inference: Towards a New Model of Measurement for Psychosocial Constructs - PubMed
Psychosocial constructs can only be assessed indirectly, and measures are typically formed by a combination of indicators that are thought to relate to the construct. Reflective and formative measurem...
110
Zach Shahn @zshahn.bsky.social · 06/08/2025
For anyone who's interested in how deck chairs should be arranged on the Titanic, we have a short note on the assumptions/rationale underlying the use of parental longevity as a proxy outcome (increasingly common) in Mendelian Randomization: arxiv.org/pdf/2508.03431
arxiv.org
110
Zach Shahn @zshahn.bsky.social · 05/07/2025
7 year old son and 3 year old daughter trade Beatles and frozen songs on car rides and he always chooses this one because it’s their longest. I’ve developed quite an appreciation for it
010