Nội dung text 9.2. Background.pdf
pacmann.io © 2022 – Pacmann AI 5 Background
pacmann.io © 2022 – Pacmann AI 6 Background Causal Diagram ● We want to remove all parts of the X / Y relationship that is related to control variable (Z) in back door path Goal: Estimate the effect of X on Y
pacmann.io © 2022 – Pacmann AI 7 Background Causal Diagram ● We can add “control variable” in linear regression model ● Control variables finds and removes variation related to the variable. Regression with Control Variable
pacmann.io © 2022 – Pacmann AI 8 Background Causal Diagram ● Regression is not the only method to close backdoor ● Making sure that there is no variation in Z also closes back doors.