Sampling properties of some econometric tests in the presense of model mis-specification
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
The objective of this thesis is to discover the consequences of applying certain diagnostic tests relating to the errors of a regression model which is mis-specified in one of several different ways. All of the tests considered are based on the null hypothesis of serial independence and all are in regular use by applied econometricians. Similarly, the non-standard conditions under which these tests are analysed are intended to represent situations which can reasonably be expected to occur in practice. It is therefore expected that the results reported in this thesis will be of direct benefit to the econometrics profession in the form of an improved understanding of the effects of mis-specification.