INTRODUCTION. Model Diagnostics and Selection

Model Diagnostics and Selection

(Ref: GP: Ch 13 )

INTRODUCTION

We can classify the practice of econometric model building in three steps: specification, estimation and evaluation.

Specification of the model is the formulation of the maintained economic hypothesis. It involves the determination of dependent (endogenous) and independent (predetermined) variables, expected sign and the size of parameters, the mathematical form (number of equations, linear or non linear form) of the model and the statement of the assumptions of the random disturbance term.

The principal of parsimony, which states that “descriptions should be kept as simple as possible until proven inadequate”, applies in model specification. If two models can serve the purpose equally well, choose the smaller one.

The choice of estimation technique of the unknown parameter depends on the nature of the relationship, its identification condition, the nature of the data and the distribution of the disturbance term.

Identification is the procedure by which we attempt to establish that the coefficients which we shall estimate by the application of some econometric technique are actually the true coefficients of the model we are interested.

Once the model has been estimated, one should evaluate the estimates to check its reliability with economic theory. This can be accomplished in two steps. Testing each equation independently and testing its applicability.

Independent test can be performed by using t-tests, F test, the Durbin Watson’s test etc., to evaluate the significance of each and every coefficient.

The estimated final model can be used either of the structural analysis, forecasting and policy evaluation. Structural analysis is the use of an estimated econometric model for the quantitative measurement of economic relationship.

Forecasting is the use of an estimated econometric model to predict quantitative values of certain variables outside the sample of data actually observed. Forecast may be the basis of action.

Policy evaluation is the use of estimated econometric model to choose between alternative policies. The selection of a most desired alternative among the various possible candidate futures would indicate which policy should be pursued.

The basic problems of econometric analysis are the model selection; the adequacy of the specification of the model and devising methods of estimation based on different error distribution.

Choosing between alternative specification of the model is the problem of model selection. Checking the adequacy of the specification of the model is the problem of specification testing and diagnostic checking.