ABSTRACT
In the previous chapters we developed the theoretical arguments underlying linear regression models. In this chapter and the next we put these ideas into practice by estimating and empirically evaluating a multiple linear regression model. In practice, the computations are carried out by the regression packages (Microfit 4/5 and EViews), which are widely available. These packages generate various diagnostic test results, which are used to evaluate the regression models in practice. We consider a number of key diagnostic tests, including those used for the detection of autocorrelation, heteroscedasticity and specification errors. Unlike conventional texts, which have traditionally dealt with these issues in several separate and disjointed chapters, and therefore give the impression to the reader that diagnostic testing and model evaluation in practice is also carried out in a similar disjointed fashion, this text takes a modern approach to empirical evaluation of the regression models, dealing with diagnostic testing procedures collectively, but in a step by step fashion. To this end, we have devoted two key chapters to the explanation of the procedures used for model evaluation in practice. This chapter deals with modelling, data requirement, estimation, criteria for model evaluation and some basic statistical procedures using a single-equation model of competitive imports to illustrate the basic idea and procedures. In the next chapter we deal with key diagnostic testing procedures used collectively in practice, to empirically evaluate single-equation regression models. Key topics
Criteria for empirical evaluation of regression models
Hypotheses testing procedures and statistical tests of significance
