Hello everyone,
For my research, I am following the methodology used in the paper "Does unmeasured confounding influence associations between the retail food environment and body mass index over time? The Coronary Artery Risk Development in Young Adults (CARDIA) study" by Rummo et al. (2017).
In this paper, the authors aim to estimate a recursive system where both W and Y are endogenous:
From my understanding, a command like: ivregress gmm B (Y W = Z) X, does not capture the recursive structure where W affects Y, and Y (in turn) affects B, while both W and Y are endogenous. It feels like what is needed would be something like:ivregres gmm B (W = Z) (Y = W Z) X, which, as far as I know, is not allowed in Stata.
What is the correct way to implement this recursive setup in Stata, where one endogenous variable is used as an instrument for another endogenous variable (i.e., a multi-step structure)?
For my research, I am following the methodology used in the paper "Does unmeasured confounding influence associations between the retail food environment and body mass index over time? The Coronary Artery Risk Development in Young Adults (CARDIA) study" by Rummo et al. (2017).
In this paper, the authors aim to estimate a recursive system where both W and Y are endogenous:
- W=aZ+bX+μ+ε
- Y=cZ+dX+eW+μ+ε
- B=fY+gX+hW+μ+ε
From my understanding, a command like: ivregress gmm B (Y W = Z) X, does not capture the recursive structure where W affects Y, and Y (in turn) affects B, while both W and Y are endogenous. It feels like what is needed would be something like:ivregres gmm B (W = Z) (Y = W Z) X, which, as far as I know, is not allowed in Stata.
What is the correct way to implement this recursive setup in Stata, where one endogenous variable is used as an instrument for another endogenous variable (i.e., a multi-step structure)?
