Hypothetical potential outcomes for six kinds of subjects, used to show why regressing the outcome on a post-treatment mediator does not recover the effect that runs through it.
Format
A data frame with 6 rows and 10 columns:
- population_share
Share of the population of this subject type, 1/6 throughout
- Z
Realized treatment assignment
- M_Z_0
Mediator if untreated
- M_Z_1
Mediator if treated
- M
Revealed mediator
- Y_Z_0
Untreated potential outcome
- Y_Z_1
Treated potential outcome
- Y
Revealed outcome
- e1
Disturbance in the mediator equation
- e3
Disturbance in the outcome equation
Source
ISPS Data Archive study D081 doi:10.60600/YU/VUH8KD
, file 42270
(GerberGreenBook_Chapter10_Table_10_1). Built by
data-raw/book_tables.R.