ABSTRACT
Much of the literature in econometrics is mainly concerned with the problems of estimation and inference from a sample of data. The properties of estimation techniques, and the quality of inferences, are heavily dependent on the correct specification of the model under consideration. For example, if a relevant variable is omitted, estimators are typically biased. If the disturbances are autocorrelated or heteroscedastic, there is typically a loss of efficiency in the estimators and the standard errors may be biased. This can lead to incorrect inferences. Thus, it is of paramount importance to invest considerable effort in checking the correctness of the specification of the model under consideration. In other words, we need to search for the best model from a set of alternative possible models using the available data. This is typically known as model selection in the econometrics literature.
