A new release of my Python library for automating estimating equations (v4.1) 🥳 This release has some minor computational improvements for clustered data, Tobit regression for censored data, and pooled logistic regression for time-to-event data

[university administration meeting] okay so here's the plan, we admit 30k more students in the next 5 years, buuuuuuuut here's the cool part, we also cut faculty and staff by 50%

I'm so excited to announce the first release of my newest #Rstats package, {adrftools}! This package facilitates estimation, visualization, and testing for the causal effect of a continuous (i.e., non-discrete) treatment. 🧵 1/10 #statssky #episky #causalinference

When does associational language make sense and when does it not? Katrina Kezios and I cover this in "How and when to use causal and associational language" with 3 suggestions for which concepts require causal language and which can be described in associational language.

New blog post about the age-period-cohort identification problem! In which, for the first time ever, I ask "What's the mechanism?" and also suggest that sometimes you may actually *not* be interested in causal inference. www.the100.ci/2026/02/13/o...

AI agent writes a PR, gets rejected, crashes out and writes a call-out blog post Absolute cinema crabby-rathbun.github.io/mjrathbun-we...

As a known HR hater, it brings me no joy to post this but I do think some papers are a bit sloppy with their criticisms of the HR (as highlighted by this nice paper) Some say the overall HR is not causal. But I don't think this is the case for reasons provided here

Cool paper from @keling-wang.com showing that 35% of randomly selected diabetes guideline recommendations use stronger causal language than the papers they cite. Such "causal jumps" are not necessarily unwarranted but, at best, lack of transparency in how causal conclusions were arrived at.

Last week I put out the v4.0 release of `delicatessen`, so I will summarize the new additions here

Last week I put out the v4.0 release of `delicatessen`, so I will summarize the new additions here

That K cliff looks like it drops off into the abyss

As an editor I often see authors fall into a trap of stressing some fancy statistical technique for estimating causal effects. Often neglected are the necessary assumptions to map the data to the estimand. @pausalz.bsky.social and I wrote about it here www.bmj.com/content/391/... #EpiSky

As I am working on a major update to `delicatessen`, a catch-up #MEstimatorMonday (50/52) about a feature I am going to add Currently, I do not automatically handle non-independent data. v4.0 is likely to change this github.com/pzivich/Deli...

This #MEstimatorMonday (49/52) let's look at beta regression. Beta regression is another regression model but for the case where the outcome Y is bounded (0,1) So, it is good for modeling proportions as outcomes. While we can also use fractional logistic regression, this is another option

I'm trying to find an example that was on Twitter back in 2021 to 2022 and hoping someone remembers better than me There was a very illustrative example of confounding by age for vaccination and COVID-19 outcomes. The vaccine looked harmful marginally, but was protective in each strata

I'm trying to find an example that was on Twitter back in 2021 to 2022 and hoping someone remembers better than me There was a very illustrative example of confounding by age for vaccination and COVID-19 outcomes. The vaccine looked harmful marginally, but was protective in each strata

You can do this right now: Think of a person who wrote a paper you love, whose work influenced or helped you, or has made your professional life better. Search up their email address. Shoot them a quick email of thanks. It means so, so much. This is a rough time of year, share some joy.

Since it's not actually Monday, I feel less guilty about this #MEstimatorMonday (48/52) which tells you how to do mediation analysis with estimating equations Let consider the effect of action A on outcome Y through mediator M

It's finals week, so let's space out on this #MEstimatorMonday (47/52). I've read a few astronomy books lately, so I was inspired to find an astrostatistics example with `delicatessen`. The following example models quasars deli.readthedocs.io/en/latest/Ex...

This not-Monday #MEstimatorMonday (46/52) is to commemorate me finally getting around to finding an example of IV analysis from the econometric literature In other words, I found Wooldridge's book and the Mroz 1987 paper deli.readthedocs.io/en/latest/Ex...

For the next catch-up-but-not-Monday #MEstimatorMonday (45/52) I am going to sketch out an argument I've been thinking about M-estimation, or the act of stacking estimating functions, has great potential to improve the statistical rigor of scientific research

For my final #MEstimatorMonday (44/52) of today, let's look at another confidence bands example This example ties in with #WorldAIDSDay as it relies on data from the AIDS Clinical Trial Group (ACTG) 175, comparing 2-drug versus 1-drug ART for prevention of disease progression among persons with HIV

For the 2nd #MEstimatorMonday (43/52) of today, let's look at those confidence bands a bit more. For this, I am going to reach outside my subject area and pontificate about cheminformatics

Turns out teaching really interferes with me staying on top of #MEstimatorMonday (42/52), so I need to do some catch up again To start, I am going to note a new release of `delicatessen` v3.2 The major addition of this release is confidence bands

What's it called when you're waiting for the bus and there are two different buses that could take you home? A doubly ro-busstop.

Going to start referring to myself as a causalologist

INCREDIBLE CURVES

I should add that I am looking to give talks on this work. If you're interested (or want to know about this work but don't have time to read the paper), please reach out I also might be able to pay my own way through some other sources, so you only need to invite me

This is one of my more out there papers. So, it is just sitting on arXiv (not submitted yet) because I am hoping to get additional feedback first So, please consider giving it a read and letting me know your thoughts arxiv.org/abs/2511.01960

I may have finally gone off the methodological deep end. A new preprint from me about some methodological ideas that have been brewing for years I'll give a bit more background on how I got here relative to the intro, and a brief overview of the paper arxiv.org/abs/2511.01960

I may have finally gone off the methodological deep end. A new preprint from me about some methodological ideas that have been brewing for years I'll give a bit more background on how I got here relative to the intro, and a brief overview of the paper arxiv.org/abs/2511.01960

I've had this on my reading list for awhile, but I finally got around to it and it is excellent It is a great introduction to bounds for missing data (that I wish I read much earlier when I was first learning about them!!) www.sciencedirect.com/science/arti...

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...

BetterBibTex's automatic sync feature in Zotero is such a great feature ... if it would work... everytime I update in Zotero, BetterBib deletes all the year and journal fields for all the entries (unless I manually re-export, thus defeating the purpose)

If you've ever found any of my work helpful, consider donating to the Python Software Foundation Learning Python during my PhD and translating everything between programming languages helped me build my understanding of causal inference. It is also why I know estimating equations as well as I do

If you've ever found any of my work helpful, consider donating to the Python Software Foundation Learning Python during my PhD and translating everything between programming languages helped me build my understanding of causal inference. It is also why I know estimating equations as well as I do

Probably a good sign that all the Meta Ray-Ban AI glasses commercials just focus on their aesthetics or celebrities wearing them, and not what you could use them for The sign of a great product!!

To close my catch-up of #MEstimatorMonday (41/52), let's talk about how I use estimating equations with synthesis models

Give it up for day 4 of my #MEstimatorMonday catch-up week (40/52). Let's go over Bespoke IVs Like proximal causal inference, this is another way to address unmeasured confounding. For this, let's go over per-protocol effects in trials

For the 3rd day of my #MEstimatorMonday catch-up week (39/52), let's talk proximal causal inference We can think about proximal causal inference as an extension of the standard identification assumptions to allow for more rich data structures. Specifically, we can account for unmeasured confounding

For the 2nd day of the #MEstimatorMonday week (38/52), I'm going to workshop a notion I've been thinking about regarding the idea of stacking estimating functions

I have been bad about keeping up with #MEstimatorMonday so this will be a week of M-estimator (so 37 to 41/52) To start, let's talk computational aspects. As you might recall from the weeks on the sandwich variance, we need to compute derivatives. One option is automatic differentiation (autodiff)

That trick? Running different regression models on the same data until you get a small P-value for that super nutrient

For years I had trouble following some of the discussion about confidence bands, but at ACIC this year @noahgreifer.bsky.social pointed me to a helpful paper So you don't have to be as perplexed as I once was, we have a new pre-print introducing the key ideas arxiv.org/abs/2510.07076

For years I had trouble following some of the discussion about confidence bands, but at ACIC this year @noahgreifer.bsky.social pointed me to a helpful paper So you don't have to be as perplexed as I once was, we have a new pre-print introducing the key ideas arxiv.org/abs/2510.07076

Aptos is such a bad font, let along one to set as the default...

Can large language models stand in for human participants? Many social scientists seem to think so, and are already using "silicon samples" in research. One problem: depending on the analytic decisions made, you can basically get these samples to show any effect you want. THREAD 🧵

I'm going to do a quick #MEstimatorMonday (36/52) to show how the pooled logit EE can be used to construct inverse probability of censoring weights (IPCW) This is based on the algorithm in my pre-print, but needs a few tweaks arxiv.org/abs/2504.13291

A well-spent $10m per year on Bill Belichick. Glad our university leadership really knows where to invest wisely

Something a bit different for #MEstimatorMonday (35/52) I'm going to take advantage of microblogging as a format and live code a new estimating equation