Measurement Error Models

This chapter discusses the use of instrumental variables for dealing with measurement error in regression covariates. Instruments are defined as observed variables that correlate with the mismeasured covariates, but do not correlate with the measurement error and with the model error. The approaches considered here are useful when little is known about the distribution of the measurement error and when validation data on the mismeasured covariates are not available. This chapter discusses the usefulness of instruments to account for measurement error in the case of linear, nonlinear, and nonparametric regression models. This chapter additionally discusses how to use instruments in empirical work when the assumption of classical measurement error is violated