Regressor diagnostics for the classical errors-in-variables model (Q583779): Difference between revisions

The author introduces the mixed linear model \(y=\beta 'X^*+\mu\) where the ``proxy variables'' \(X=X^*+\epsilon\) are subject to the measurement errors. A series of diagnostics are developed to prove the sensitivity of regression coefficient estimates to measurement errors in each regressor (large-sample-properties of the classical errors-in-variables model). Algorithms to compute the diagnostics are provided, and an example is used for illustrations.