https://www.jstage.jst.go.jp/article/jea/30/9/30_JE20200226/_article

PMC

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7429147/ 

Highlights

• Marginal structural models (MSMs) should be distinguished from inverse probability weighting.
• MSM shows prespecified assumptions on causal estimands, while an exposure probability model is an imposed restriction on observed distribution.
• As MSM and exposure probability model are used for different purposes, misspecification of these models would lead to biases in different ways.
• Model specifications of MSMs and exposure probability models raise different challenges in real data analysis.
• G-formula, which shares identifiability assumptions with inverse probability weighting, can be used to fit MSMs only when the models are saturated.

 Selected Result