🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Medicine is the perfect proving ground for developing AI/ML for causal inference (like estimating treatment effect from non-randomized observational data). Because can validate against our randomized trials, that other fields really don't have. Will see lots of progress ~soon.

For anyone interested in stats/ML in oncology, please check out the 8th annual Stat4Onc symposium, happening at Stanford in May. I'll be speaking in the precision medicine session. stat4onc.org/events/stat4on…

A great blog post from Jessica Hullman on why "dumb" statistical models tend to outperform doctors in diagnostic tasks and how much it matters. The references are very cool too. statmodeling.stat.columbia.edu/2025/04/18/dum…

The billing system is an opaque mess. It's super stressful for patients too. A patient told me this week they were in an imaging trial at an unnamed center (no standard of care procedures, just research) and were accidentally billed $50K. Now they are hesitant about research :(

OpenEvidence (OpenEvidence) is really impressive. Kind of like UpToDate but if it just printed out the answer. Quality is high too. I love UpToDate but this feels a bit like Google with its magic algorithm going up against Yahoo with its human-curated links list in the late 90s.

Well this took a bit longer than expected, but I think it's finally ready! Tweet thread coming soon. Kind of anxious but hoping that some people will find it useful.

Congrats, we really needed a win for this patient population (mainly oral cavity pts) with high recurrence rates after standard treatment. 8% OS benefit at 2yr for CPS >= 10 group (with a CI that doesn't cross 1) adds to the strong EFS results.

I hate to say it but this CAPTCHA was actually pretty fun to complete.

For a reminder of why causal inference with retrospective data is tricky, check out Jon Mellon's paper on instrumental variable analysis. Featuring this beautiful and horrifying DAG showing how papers often oversimplify causal relationships. onlinelibrary.wiley.com/doi/10.1111/aj…

The peer-reviewed version of expansion in situ genome sequencing is now out in Science! The news is bittersweet – when we first revealed this last September, I never guessed it would be my final paper in academia, but a lot has changed. A few parting thoughts: