Very well said. Re (2), a version of this is in the Korn and Graubard "Analysis of Health Surveys" textbook that explains how, under misspecification (and all models are somewhat misspecified), the weighted (population) fit is what we want rather than the sample fit.

Skepticism toward vax and modern medicine is fairly widespread publicly as I’ve seen, and is tied up with antipathy toward “experts” (“do your own research” a la Aaron Rodgers), which universities embody. There is the woke issue sure, but there is also antipathy toward expertise to contend with.

Let’s take the area where there has been the most heat: vaccinations. The MRNA vax researcher doesn’t need to change what they do. It’s a problem of educating the public that this is useful stuff. The question is how to do that effectively.

Seems that a field that has taken an especially large hit in the current context is biomedical sciences, and this because, it seems, MAGA rejected the expert guidance on COVID. These rules leave many unanswered questions for that field, no?

I agree with comments here emphasizing Trump’s vengeance on the “medical establishment” after feeling that it conspired during COVID to unseat him. Would be nice if nurses and doctors generated a groundswell of alarm in their personal social networks about the dangers.

Montiel Olea has a few really nice papers on design from a decision theoretic perspective.

Anna Wilke and I have a paper (soon on Arxiv) that I can send on design diversification from minmax regret perspective. We were inspired by Manski and papers on experimental design like Banerjee et al. and Azevedo et al.

www.aeaweb.org/articles?id=...

www.journals.uchicago.edu/doi/abs/10.1...

Isn't this "generalization"?

(The tortured fit of a global polynomial)

*cited (not cute)

We had a wee contribution on this for randomized experiments (cute by Abadie et al.). The idea extends by analogy to observational studies: cyrussamii.com/wp-content/u...

One observes only a sample of the potential outcomes. Sampling error based standard errors provide for conservative inference for such uncertainty. Surprisingly, this was not widely known until fairly recently. (Cf the Abadie et al paper and refs therein.)

Then, I love this paper by Allen et al. although I suspect you have it already :-)

www.science.org/doi/10.1126/...

This paper uses observational data to extend results from a field experiment: www.cambridge.org/core/journal...