Diagnostic Testing: An Application to the Demand for M1

Abstract

Diagnostic or specification tests are typically used as a means of indicating model inadequacy or failure. For example in the case of a linear regression model which is estimated by ordinary least squares (OLS), a series of diagnostic tests could be used to indicate whether any of the assumptions required for OLS to be the best linear unbiased estimator (BLUE) appear to be violated. These assumptions include a serially uncorrelated and homoscedastic error term, absence of correlation between the error term and the regressors and correct specification of the conditional mean function, i.e. no omitted variables and appropriate functional form.