/blog/2021/01/15/msm-gee-multilevel/

[andrewheiss]

Marginal structural models for panel data with GEE and multilevel models | Andrew Heiss
https://www.andrewheiss.com/blog/2021/01/15/msm-gee-multilevel/

[andrewheiss]

Originally posted by Felix Thoemmes on 2021-01-21 00:57:52

Hi Andrew - glad that the paper was useful. You are correct, there are
two typos. You are the second person to catch them (somebody emailed me
a while ago). I wish I wouldn't have made those typos, because it trips
people up, and hinders understanding. Too late now. And sorry that the R
code was only in the PDF - they are a pain to copy and paste out of PDF
papers. Feel free to email me next time you need some of my code (when
Cornell redesigned their website they broke the link to the code repo
that is in the paper).
Anyway, this was a great blog post, and I enjoyed it a great deal!

Minor comment:
I think your DAG does not include the auto-regressive paths that we
included in our data-generating model, and we used the survey package to
get robust standard errors - even though I like your approach of using
geemodel better.