Top 3 Model Specification Tests in Econometrics

Ramsey’s Reset Test

Non-nested F test or Encompassing F test

Davidson Mackinnon J test

This test is used to test mis-specification of models. In this test we run regression on model 1 and obtain estimated value of regressand and then use the square and cube of  estimated value as additional regressors in model 1 and name it model 2. Now with the help of coefficient of determination of both models, run a f test. If the f-values are significant then accept the hypothesis that the model is misspecified.

This test is used to choose a better model between two competing models. In this method model 1 and model 2 nests or encompasses in a single model (but model 1 is not nested in model 2 and vice versa) and then we use F-test to choose which model to consider. If the coefficient of explanatory variables becomes insignificant then we choose model 1.

This test is an alternative to non-nested F test. In this test we calculate the estimated value of regresand (y) from model 1 and use it as an additional regressor in model 2 then apply t- test. If the coefficient of additional regressor is insignificant, we accept model 2.