Timeline for Can we perform matching on post-treatment variables?
Current License: CC BY-SA 3.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 27, 2023 at 21:05 | comment | added | RobertF | Depending on the duration of the post period and when observations are measured, it's ambiguous if the outcome $Y$ is causing the post-treatment variable $X_{post}$ or $X_{post}$ is causing $Y$ - or both! In either case, the pre-treatment measure of the variable, $X_{pre}$, is also causally linked to $X_{post}$ so that $X_{post}$ is a collider which we want to avoid conditioning on. | |
| Mar 31, 2018 at 0:00 | history | edited | usεr11852 | CC BY-SA 3.0 |
Added explaination.
|
| Mar 25, 2018 at 1:26 | comment | added | usεr11852 | We do not directly do anything with the post-treatment variables. That being said, they are useful, we can use these post-treatment variables as additional descriptors in the study or maybe use them to propose new hypotheses. We should be very careful not to use the post-treatment variables to immediately test new hypotheses though because that would amount to post-hoc theorising which is another problem in itself. | |
| Mar 24, 2018 at 19:24 | comment | added | msmazh | So, then what should I do with those post-treatment variables? How should I control for them? Say I count # of tweets before and after treatment. Then matched users according to # of tweets before treatment. Then how should I account for # of tweets after treatment? | |
| Mar 24, 2018 at 1:17 | vote | accept | msmazh | ||
| Mar 23, 2018 at 23:52 | history | edited | usεr11852 | CC BY-SA 3.0 |
deleted 1 character in body
|
| Mar 23, 2018 at 23:38 | history | answered | usεr11852 | CC BY-SA 3.0 |