Instrumental regression techniques have played a key role in both the theory and practice of econometrics. There is a growing literature on the nonparametric treatment of this topic. The basic framework involves a response variable denoted Y and a vector of explanatory variables denoted Z. The variables in Z can be endogenous or exogenous and the relationship of interest is given by Y = φ(Z) + U where φ(·) is the function of interest but will not, in the present setting, coincide with the conditional expectation function E(Y|Z) hence requires special treatment (the exception being when all variables contained in Z are in fact exogenous). This talk outlines recent advances in nonparametric instrumental regression methods and discuss potential application in the health arena.

Recent Advances in Nonparametric Instrumental Regression (.pdf file)

By: Jeffrey S. Racine (McMaster University)
2014 Annual Health Econometrics Workshop (AHEW)

R code to replicate examples in slides is available for download. Users are required to:

a) Install R (if not already installed)
b) install the np package via the R command install.packages (“np”)

Note: Code is numerically intensive and can take 1/2 an hour on a modern desktop/laptop.

R Code for Slides (.zip file)