Tuesday, June 13, 2006

Statistical Problems: Intermediate Outcomes

Via Newmark's Door:
Andrew Gelman presents a nice example of, and briefly discusses, a statistical issue that should get more attention: "the fallacy of controlling for an intermediate outcome". (Economist Mark Killingsworth--American Economic Review, May 1993--labels it "the included-variable problem".)
The fallacy of intermediate outcomes, or what has sometimes been called the included varaible. If one is interested in income in 1991, but they use total education measured in 1993 as a RHS variable, then thye have commited this mistake.

For more details see Andrew Gelman at the must read blog: Statistical Modeling, Causal Inference, and Social Science.

Keywords: ECO307, BUS735

No comments: