Further Results on the Errors‑in‑the‑Variables Problem

Notes on the omission of two important aspects of the errors-in-the-variables problem--the effects on the power of the [t]-statistic test when the explanatory variables in an OLS regression contain measurement errors, and the case in which there is measurement error in more than one explanatory variable. This paper is concerned with the effects of measurement errors in one or more explanatory variables on the asymptotic bias of the coefficients and the [t]-test. The discussion progresses from examination of a simple regression model with one explanatory variable, to a model with two explanatory variables, and finally, an expansion to the [n]-variable model by way of partitioning. (October 1971)