RT @StatMLPapers: Generalized Venn and Venn-Abers Calibration with Applications in Conformal Prediction

@CAUSALab @karolinskainst @GelovenNan @UniLeiden Amazing as always @GelovenNan I think that one cognitive barrier for IPTW is the constant updating for treatment weights, the same way as Blanche did for censoring and AUCs. This procedure makes the performance metrics seem not as agnostic as the conventional ones.

RT @MaartenvSmeden: #4: Abuse of causal language tops many peoples list (including mine) of most annoying things in the scientific literatu…

RT @f2harrell: #Statistics thought of the day: Semiparametric ordinal regression models work spectacularly well for continuous dependent va…

RT @CAUSALab: 2 weeks away! @CAUSALab Methods Series @karolinskainst resumes w/ @GelovenNan (@UniLeiden) presentation on predictions under…

@f2harrell @TopfHNS @TassoneHNS @sindhurasridhar You might think about the direction of the hazards for fine-gray, but not about the magnitude. Either way consider death as censoring is an extremely weird thing conceptually. You basically claim that ghosts might have heart-attack (primary event) but we just mis the data 👻

RT @f2harrell: @TopfHNS @TassoneHNS @sindhurasridhar Only for assessing some mechanism, never for judging treatment effectiveness. Even mo…

RT @VickersBiostats: What is a good medical choice? A critique of some current dogmas in medical decision making

📢 Webinar announcement! Really excited to host @smueller for a webinar about personalised decision making! 🗓️ 8.1 This Wednesday 20:00 gmt+2 Subscription here: 👇🏼

I wasn't familiar with the term Eikosograms, thanks @soboleffspaces! I tend to think they might be useful for data exploration, but not sure about data storytelling. This is a nice tutorial 👇🏼

@soboleffspaces I'm also working on benefit bounds viz similar to mosaic plots. Thats just a poc for now: Would love to hear some feedback or any ideas

For great Educational Causal Inference material you should check Causality with Boris:

Just rerendered my presentation about decision curve analysis with quarto! I would like to thank @VickersBiostats for his help @quarto_pub

Three myths about risk threshold for prediction models by @Laurewts 🧞

@soboleffspaces @yudapearl Thanks Boris! I like the interaction framework, but I'm not aure if that's always the common case. For CVD prediction we think a lot about Statins vs No-Statins. Also What about sport activity? Or vaccines? Just one treatment is considered and we still lack the term for that

RT @soboleffspaces: I was thinking about your question a bit more. @yudapearl causal epistemology addresses the treatment of existing con…