Research Design Flashcards
ATE
E(y1) - E(y0)
NATE
E(Y1 I D=1) - E(Y0 I D=0)
ATE
π x ATT - (1-π) x ATC
ATT
E(Y1 I D=1) - E(Y0 I D=1)
ATC
E(Y1 I D=0) - E(Y0 I D=0)
Total bias
NATE - ATE
BB
E(Y0 I D=1) - E(Y0 I D=0)
DTEB
(1-π) x [[E(Y1 I D=1) - E(Y0 I D=1)] - [E(Y1 I D=0) - E(Y0 I D=0)]]
Internal validity
able to identify the causal effect in the study sample
external validity
ability to generalize the results to other contexts
confounder
when Z points outwards (do D and Y or other variables)
mediator
when one arrow points into Z and another out of Z (on the path between D and Y or other variables)
collider
when arrows point to Z, both variables effect Z (path closed)
direct path
when there is a causal open path between D and Y
non-causal open path
when there is an open path between D and Y but not causal (backdoor paths)