Things to ask / discuss with HCM authors

[adamrupe]

For instance, it is not stated explicitly (e.g. in pseudo code) when variables should be augmented or marginalized (although we are piecing this together from the text and examples).

• Clarification of J.1: General form of augmentation variables. There is a typo at the end; I'm pretty sure they mean Figure A3j instead of A3k. They also have q(a; do(x)), which is not consistent with the 'a' being the treatment variable throughout the rest of the paper. Similarly, why does (69) have 'do(x^(script R))'? This also does not seem consistent with the rest of the paper.
• What are the specific causal queries for each example in Figure A3? It appears that some are conditional interventions (i.e. those with 'a|x'), but it is not explicitly stated.
• Why is Q^w augmented in Figure A3 (q)? Is this necessary for identification (maybe front door adjustment)?
• Is there a general algorithm for identifying when a variable needs to be augmented to enable identification?