search

r-causal / causal-inference-in-R

Causal Inference in R: A book!
https://www.r-causal.org/
165 stars 42 forks source link

First g computation sections #225

Closed tgerke closed 3 months ago

tgerke commented 6 months ago

The broader chapter still in progress, but would be good to get some feedback on this presentation ordering (i.e. a preference to show Monte Carlo first, with the "data stacking" technique introduced later as a special case)

malcolmbarrett commented 6 months ago

Ok, I think I'm on board with this because I think I see the pedagogical advantage. Maybe the data stacking can even just be a longer callout box, e.g. for these simple cases you can use this one weird trick

I think we should address the question of sample size, like how does the sample from the model relate to the number for the monte carlo and the ultimate variance, e.g. the one from the bootstrap or whatever? I think this might relate to how g-comp can technically be more efficient while still being accurate but I don't remember the nitty gritty of our conversation about this

tgerke commented 6 months ago

Cool, good call re: sample size. I was planning to be more intentional about this is the subsequent section that uses bootstrap for CIs etc. Will remember to circle back to the initial section to pick a resample size that's more deliberate. I have a ref for this somewhere, stay tuned :)