once again learning the painful lesson that i need to pick a color scheme and set the correct page size before polishing my tikz figure 🥲
07.03.2026 00:00
👍 4
🔁 0
💬 1
📌 0
once again learning the painful lesson that i need to pick a color scheme and set the correct page size before polishing my tikz figure 🥲
i'm intrigued by all the design-based material!
Do epi folks ever use IV for per protocol designs or is the LATE just rarely of interest?
I think an underappreciated benefit of vectorizing is that you often use a more functional programming pattern and there's less context switching
addendum: this is about what happens at the knowledge frontier, where things are more nebulous. away from this frontier i think more is possible, pending social trust in the extent body of work
yeah i think the funding argument (at the social level, not the individual grant level) is more along overall roi lines, like www.unitedformedicalresearch.org/statements/u...
my personal impression is that knowledge as a scientific product is a bit optimistic (in terms of what we actually produce)
i rather like the idea of shared understanding as the product/output of scientific processes though
New on the arxiv:
“Graphs are maximally expressive for higher-order interactions”
arxiv.org/abs/2602.16937
We clarify central misconceptions in the recent literature on "higher-order networks".
w/ @piratepeel.bsky.social , @manlius.bsky.social, and @thilogross.bsky.social
Explainer 🧵: 1/N
Thanks!
Is there an #rstats package to visualize how many observations get dropped by each exclusion criteria?
Looking for a nice visual way to summarize a sample
New paper! The Linear Representation Hypothesis is a powerful intuition for how language models work, but lacks formalization. We give a mathematical framework in which we can ask and answer a basic question: how many features can be stored under the hypothesis? 🧵 arxiv.org/abs/2602.11246
"Ten simple rules for teaching data science": arxiv.org/abs/2602.02874
A new preprint by @minecr.bsky.social and myself. We'd love any feedback!
All of this code is from the ccesMRPprep package, but this just strips things down to the bare minimum to pull census crosstabs so it's a little easier to understand, or if the CCES part of the data prep isn't relevant to you
Quick gist to pull census crosstabs:
- age x sex x educ
- age x sex x race
- sex x race x educ
to use during post-stratification
gist.github.com/alexpghayes/...
if i'm understanding correctly, this repo has already-pulled poststrat tables, and then fits models and does poststratification?
cc @awmercer.bsky.social @gelliottmorris.com
Question for the pollsters: are there any public repos with census data/code to pull census data for re-weighting surveys with Mr. P?
At long last @zotero.org has an Android app!
hmmm something like E[max coreness | author]? napkin sketches suggest this is ~scaled max degree correction parameter of author in DC-SBMs
can you say more about how you're running this? i managed to connect claude code to this mcp server but i'm much more interested in getting wolfram connected to a reasoning model
can you share any references about these more effective strategies? i'm interested in reading more!
Solidarity with everyone striking against ICE in MN and elsewhere
If you'd like to do something: I'll match up to $100 in donations to any of the organizations listed on www.standwithminnesota.com
{broom} has a lot of early tidyverse baggage! It's great that other packages have smoothed things out!
I'm looking forward to teaching a 4-hour short course on Bayesian sensitivity analysis methods at the American Causal Inference Conference (ACIC) 2026!
Register here: sci-info.org/annual-meeti...
In a shocking turns of events, making a small change to my simulations did in fact fix everything
Also it's not simple! I think cognitive aspects of learning the new semi-parametric methods is a fairly serious constraint for applied use
Yeah I thought epi used sequential G things? Also, I'm not sure I think of front door as all that morally different from IV
You get a free license through UW
@karlrohe.bsky.social both this and the mathematica mcp bit higher up the thread
I've pulled out of several projects but far before any manuscript writing began