Bounding the effects of measurement error in regressions involving dichotomous variables (Q1117646): Difference between revisions

In a multiple regression involving dichotomous independent variables, the effects of measurement error in the dichotomous variables is analyzed. Unlike the case of classical (i.e., white noise) measurement error, the measurement error in a dichotomous variable must be correlated with the true variable and any other variables correlated with the true variable. Nonetheless, it is shown that the results for the classical errors-in- variables model can be used to analyze the implications of measurement error in dichotomous variables. It is shown how the bounding procedure developed by the author and \textit{E. E. Leamer} [Econometrica 52, 163-183 (1984; Zbl 0526.90030)] for the classical errors-in-variables model can be adapted to bound the coefficients when the regression contains mismeasured dichotomous variables. The classical results are also used to analyze the inconsistency in least-squares regression coefficient estimators computed from data on the mismeasured variables.