Handelgroup Bayesian Censoring Project

We will keep our notes and code on dealing with censored variables in Bayesian models in this repo. My initial idea for this is that we can basically treat each worked out example or section that we write up as a chapter in this quarto book, and then when we are ready to pull everything together into the manuscript that should make it easier, and we could potentially make this website available with some edits.

TODO

• Basic description of censored data problems (intro.qmd)
• Make a glossary of terminology so we can all be on the same page with language.
• Description of how censored data is dealt with in the frequentist paradigm and why this can be difficult. Maybe a walkthrough of methods with certain packages like survReg, survival, NADA2, AER.
• Description of how Bayesian models can handle censoring and a brief conceptual introduction to our examples, maybe a brief description of the two different methods. This could be expanded if we learn enough to compare them.
• A few worked examples using Stan and potentially brms to show how these models are set up and executed. don’t check this off until a FULL FIRST DRAFT OF THAT PAGE IS DONE!
  • Basic examples that explain concepts
    • Censored predictor
    • Censored outcome
    • Both
    • Mix of censored and uncensored / multiple predictors
    • Interval censoring separately because I think it’s harder
  • Data-motivated examples
    • Binary outcome with censored predictor (norovirus project)
    • Interval censored outcome with LoD (all HAI projects)
  • brms censored outcome formula tutorial, extract stan code, and explain what the stan code is doing
• Add the little github link to the top corner
• Create a zotero group library and add the link to this page

Guide to collaborative work

• The Zotero group library containing the references for this project is here. Please request membership (and email/IM Zane) if you want to add/edit references.
• At some point we probably need to use GH actions to automate building, because that is the main step of the process where conflicts occur.
• For now, we should try to work in separate documents to prevent conflicts.
• Try to remember not to commit changes to the docs folder! We should only rebuild the site when we know all work has committed and everyone who is working on the project is ready to do that. Otherwise we will get annoying merge conflicts that are very annoying to fix.

Current contents

• Zane’s stan example extending the first example from that StackExchange post, specifically trying to extend the answer by Tom Minka.
  • fix this code and make data so we can run this example.
  • write up an explanation of the motivation and what’s going on here.
  • download an HTML backup of this web page and embed it somewhere here in case anything ever happens to it or for when SE is down.
  • frank harrell comment method?
• above post, getting the method by Bjorn running
• Tom method for censored outcome (first method in Stan docs)
• Bjorn method for censored outcome (2nd method in Stan docs)