Arthur Albuquerque

wat

@f2harrell.bsky.social how could one implement Bayesian imputation previously applied in model development? Report the posterior draws of each coefficient from imputation models?

Forgot to mention the biggest shortcoming: not having the outcome available

Problem with MI is when one does not have the dev/validation data available when deploying the model, which is not uncommon

What method would you use if planning to deploy a CPM allowing for missing data for the end-user?

Similar arguments by Riley and Martin in the preprint I mentioned

but doesn’t frequentist single regression imputation also perform better if Y is omitted?

Compatibility of Missing Data Handling Methods across the Stages of Producing Clinical Prediction Models Missing data is a challenge when developing, validating and deploying clinical prediction models (CPMs). Traditionally, decisions concerning missing data handling during CPM development and validation...

I believe they will discuss this enlightening paper: arxiv.org/abs/2504.06799

in which they recommend single regression imputation when ones plan to apply a CPM and wants to allow for missing data at deployment.

What would be the advantage of Bayes in this case?

A "Mr. McMahon reaction meme" where McMahon shows increasing levels of amazement/excitement, with levels "I^2", "tau^2", and "prediction interval", and "predictive distribution".

Great talk. In CPMs using LASSO, would you recommend reporting the proportion of how often each predictor was selected by LASSO in bootstrap samples?

thanks! just did

informative prior for tau could help a bit

COULD EVERYONE PLEASE STOP USING RANDOM EFFECTS META-ANALYSIS WHEN COMBINING 3 OR 4 TRIALS? AND COULD REVIEWERS STOP DEMANDING IT?

Nephrologists vs. Intensivists: Cumulative ordinal model Nephrologists vs. Intensivists: Cumulative ordinal model - gist:d6b0e3fec5f9a70e86dcc7dcdd7a4c19

code gist.github.com/arthur-albuq...

Couldn't hold myself, had to fit a #Bayesian cumulative ordinal model to this interesting data:

Canadian intensivists and nephrologists don't seem to disagree much!

Always happy to share Brazilian music! THE best 🇧🇷

Acenda o farol YouTube video by Tim Maia - Topic

this is one has been on hype lately in Brazil. His best bass line! youtu.be/tPp9zkjfj7Q?...

O Descobridor Dos Sete Mares YouTube video by Tim Maia - Topic

and this is his funkiest one! youtu.be/PAUlCK8kuGU?...

Gostava Tanto De Você YouTube video by Tim Maia - Topic

youtu.be/cxSzri346W0?... check the lyrics! love this one

Evaluating the risk of ovarian cancer before surgery using the ADNEX model to differentiate between benign, borderline, early and advanced stage invasive, and secondary metastatic tumours: prospective... Objectives To develop a risk prediction model to preoperatively discriminate between benign, borderline, stage I invasive, stage II-IV invasive, and secondary metastatic ovarian tumours. Design Obser...

Tough one. There is a famous clinical prediction model with such an outcome, if that’s of any help: www.bmj.com/content/349/... by @benvancalster.bsky.social

Added a statistical analysis tab to explore Bayesian partial pooling visualization (cc @rmcelreath.bsky.social). Added some model explanations too.

Data viz heavily inspired by @benmoran.bsky.social's beautiful bayesfoRest package

Amazing work. I was particularly interested in visualizations for studies on dexmedetomidine for investigating dosage, timing, risk of bias.

Vibe-coded this app: arthur-albuquerque.github.io/dexmedetomid... Example visualization attached.

code: github.com/arthur-albuq...

I'd love some feedback

I will investigate the best approach and do it! thanks again

I meant for your “Create symlinks” section.

For Roo, I believe the reference directory for skills is

• ~/.roo/skills

This is so cool. Thank you!

Please add suport to Google’s Antigravitiy, Positron Assistant and VScode’s extension Roo Code.

These are more scientific-oriented softwares

@matthewluney.bsky.social et al really amazing work, congratulations.
Quick Q: you stated that "Moderate certainty evidence indicates that dexmedetomidine is effective". However, CINeMA showed low confidence rating for dexmedetomidine vs. placebo (Table S22).

Why did you reference it as "moderate"?

creating a custom_family with a categorical-like response (e.g., Dirichlet-multinomial) · Issue #1728 · paul-buerkner/brms I've been attempting to create a custom_family for the Dirichlet-multinomial in brms v2.22.0, following the approach set out in the custom family vignette. However, I'm running into difficulties. T...

github.com/paul-buerkne...

Important research question!

What do you think about multinomial-dirichlet? It’s been implemented in brms

Isn’t this formula utility very limited?