Some Implications of Random Measurement Error for Causal Inferences

Random measurement errors can lead to faulty inferences by introducing additional unknowns. This makes pausible a number of alternative explanations that should be explicitly noted. One can study the implications of measurement errors by adding measurement error components to causal models. This approach is used in five specific situations: (1) where one is testing for spuriousness or interpreting by means of intervening variables; (2) where independent variables are highly correlated; (3) where there is a developmental sequence; (4) in factor analyses; and (5) where interaction might be inferred.