Chapter 39 Controls

This section follows (Cinelli, Forney, and Pearl 2022) and code

library(dagitty)
library(ggdag)

Traditional literature usually considers adding additional control variables is harmless to analysis.

More specifically, this problem is most prevalent in the review process. Reviewers only ask authors to add more variables to “control” for such variable, which can be asked with only limited rationale. Rarely ever you will see a reviewer asks an author to remove some variables to see the behavior of the variable of interest (This is also related to Coefficient stability).

However, adding more controls is only good in limited cases.

References

Cinelli, Carlos, Andrew Forney, and Judea Pearl. 2022. “A Crash Course in Good and Bad Controls.” Sociological Methods & Research, 00491241221099552.