Really enjoying the celebration of 50 years of the Cox model.
Turned up an hour early to do some work in the coffee shop, bumped into Bhramar Mukherjee who was there for the same reason. Neither of us did any work but had a great chat!
Some really good talks so far: Nancy Reid, Per Kragh Anderson, and now Hein Putter. It’s quite tricky to come up with questions for ‘review’ rather than ‘ideas’ talks.
Yess we’re talking about sufficiency. Love a bit of sufficiency. Also ancillarity.
This is is Heather Battey of Imperial College London, talking mainly about how Cox brought clarity to many of Fisher’s ideas.
Prof Bhramar Mukherjee is wearing the lasso/ridge dress!
What a presenter.
Rhian Daniel (Cardiff) presenting the most enigmatic title of the day: Regression by composition. Really looking forward to this one.
I’ve made it!
Rhian finally made the answer to my question clear, a year later.
@Tim_P_Morris is it 1, 2, or both?!
@Tim_P_Morris thinking a little more about it, the (non-)collapsibility difference probably has to do with the relationship between PH and AFT for the Weibull:
\beta_{PH} = -\gamma \beta_{AFT}
@Tim_P_Morris also just randomly came across this paper
https://pubmed.ncbi.nlm.nih.gov/36271175/
@PausalZ Interesting. Another for my 'I'll never fully understand collapsibility' list.
https://www.jstor.org/stable/26511790#metadata_info_tab_contents
@PausalZ Yes I think that's a helpful point
@PausalZ HR is still noncollapsible and acceleration is collapsible. The explanation was that when you collapse it’s no longer Weibull, and it seemed much clearer with Rhian’s pictures of the compositions.