Congratulations Dr. Keya Joshi for a successful defense yesterday of her PhD thesis Vaccines: Populations, Individuals, and Models.

Worried about test data being used in training? The LLM world is going through a data contamination crisis. Here's us trying to do something about it: Paper: arxiv.org/abs/2305.10160 Blog: medium.com/@alonjacovi/st… w\ Avi Caciularu omer goldman (((ل()(ل() 'yoav))))👾

stanforddaily.com/2023/05/25/sta…. An "absolute honor" to work for a man whose political strategy is to use LGBTQ+ people, immigrants, and universities as punching bags to widen divisions and appeal to the worst in our country. Nice.

Important points in this paper, really worth reading and acting on as a form of pandemic preparedness.

#SER2023 is officially here, and predictably (R^2=.99) the first poster session tonight was a massive Harvard Epidemiology reunion! 📸: ~Clustering~ in front of @KerollosWanis, Arin Madenci & my (@ChelsMessinger) poster on causality in surgical RCTs w/ longitudinal components

You know they may be destroying science, but somedays my ego is really here for the emails from predatory journals reminding me of my "outstanding reputation and contribution in my field" and telling me my work is "first-class and of the highest quality"

This seems really nicely done and has wide applicability beyond education interventions. For instance it fits with observations about variation in implementation quality and heterogenous effects in a violence reduction program here: pnas.org/doi/10.1073/pn…

“Implementation matters: Generalizing treatment effects in [insert development field here]” So obvious, but nice seeing it incorporated formally into analysis!

🆕 DP: We experiment w/ interviewer vs. anonymous collection of violence measures in Senegal 🇸🇳 --> Anonymous [self-administered] methods ⬆️ reporting by 4-12 pp depending on type of violence Joint w/ Malick Dione Dr Fatma Lamesse Diédhiou IFPRI IFPRI_Africa ebrary.ifpri.org/digital/collec…

Two things can be true you can’t learn everything from an RCT and yet for most things it would be wayyy better if we had one.

If anyone has some spare codes for the other site I would like one plz 🙏

Ideas for epidemiological study design and interpretation and for how modelers should use these findings. Led by Keya Joshi with Rebecca Kahn and Christopher Boyer