I found it! See chapter 11 in Statistical Rethinking. Always check that first.

Or wait is that just that you need to be careful when doing Bayesian cross validation about what unit of analysis you are using: the individual outcome vs a unique combination of predictor values?

I read somewhere about how logistic regression with binary individual outcomes vs count data can lead to different posteriors if you fit a Bayesian model. Anybody remember in detail what I seem to remember?

But if you mean the outcome, then I am agnostic about what type of outcome it is as long as I've got options for how to consistently define treatment.

Complex mixture of the two. Also interested in how to defining a treatment that is basically a projection of some complex unstructured or semi-structured data (say, the terms in a contract) into a lower-dimensional space.

HELP REQUESTED FROM 📈📉 #EconSky: Looking for causal inference papers where someone has cleverly encoded a complex treatment (say, the terms of a contract) and then estimated a variety of causal estimands on the back end of the analysis that are all interpretable to a human being.

Is there a reporting standard for heterogeneous treatment effect estimation, akin to the AGReMA standard to mediation analysis?

So we have MMA fighters name dropping von Mises now…

for any applied econ folks treating rainfall as exogenous... 📈📉

KHkellyhereid.bsky.socialMar 10, 2024 2:07am

New climate finance scam just dropped: Colorado ranchers scam index-based drought insurance for $3M by putting pie tins over rain gauges. HT @bryanwx.bsky.social

Colorado ranchers sentenced after tampering with rain gauges to increase crop subsidies

Two southeastern Colorado ranch owners were recently sentenced to pay $6.6 million to resolve federal charges that they damaged or altered rain gauges in an effort to get paid for worsening drought co...

The dragon gets it. 📉📈