@Prof_BearB @conjugateprior @lauretig @rlmcelreath I've been very bad / busy and not organised to receive @psmaldino 's new book, but maybe he knows whether this is in it?
Follow
@j2bryson @Prof_BearB @conjugateprior @lauretig @rlmcelreath
It’s not. I don’t really see how CI would work for a purely mathematical model, since the relationships are all spelled out in the equations. No?
@conjugateprior @psmaldino @j2bryson @Prof_BearB @lauretig Sorry to reply to such a huge canoe. Here's a paper on counterfacturals in simulations that helped me think about these things. Lots of causal implications are non-obvious I think. https://doi.org/10.1098/rstb.2018.0279
@psmaldino @j2bryson @Prof_BearB @lauretig @rlmcelreath They are. But not all their implications, most of which are counterfactual.
As a comparison, a garden variety regression model is purely mathematical model but it's still illuminating (and non-obvious) to know, say marginal and conditional effects.
Even if these are only derived as observable implications I assume they could be useful when aligning the model to data (unless everyone uses ABC for that these days, I don't really keep up)