4.3 Test for Dependence

One must always remember that an apparent problem, such as residual autocorrelation, could have other causes than error autocorrelation, so it is worth investigating alternative sources, such as incorrect functional form, or parameter non-constancy (discussed in section 4.4). (Hendry and Nielsen 2007)

4.3.1 for Autoregressively Dependent Innovations

4.3.2 for Temporally Dependent Innovation Variances

4.3.3 Improve the Model

A failure of a test for temporal independence of the innovations can often be remedied by reconsidering the specification of the model. (Hendry and Nielsen 2007)

References

Hendry, David F, and Bent Nielsen. 2007. Econometric Modeling: A Likelihood Approach. Princeton University Press.