Uncertainty in Artificial Intelligence
First Name   Last Name   Password   Forgot Password   Log in!
    Proceedings   Proceeding details   Article details         Authors         Search    
On a Class of Bias-Amplifying Variables that Endanger Effect Estimates
Judea Pearl
Abstract:
This note deals with a class of variables that, if conditioned on, tends to amplify confound- ing bias in the analysis of causal effects. This class, independently discovered by Bhat- tacharya and Vogt (2007) and Wooldridge (2009), includes instrumental variables and variables that have greater influence on treat- ment selection than on the outcome. We offer a simple derivation and an intuitive explana- tion of this phenomenon and then extend the analysis to non linear models. We show that: 1. the bias-amplifying potential of instru- mental variables extends over to non- linear models, though not as sweepingly as in linear models; 2. in non-linear models, conditioning on in- strumental variables may introduce new bias where none existed before; 3. in both linear and non-linear models, in- strumental variables have no effect on selection-induced bias.
Keywords:
Pages: 417-424
PS Link:
PDF Link: /papers/10/p417-pearl.pdf
BibTex:
@INPROCEEDINGS{Pearl10,
AUTHOR = "Judea Pearl ",
TITLE = "On a Class of Bias-Amplifying Variables that Endanger Effect Estimates",
BOOKTITLE = "Proceedings of the Twenty-Sixth Conference Annual Conference on Uncertainty in Artificial Intelligence (UAI-10)",
PUBLISHER = "AUAI Press",
ADDRESS = "Corvallis, Oregon",
YEAR = "2010",
PAGES = "417--424"
}


hosted by DSL   •   site info   •   help