The study of the effect of the violations of the model assumptions on the parameter of interest is called sensitivity analysis. Chapter 2 of this thesis defines the sensitivity statistic and confronts sensitivity analysis with diagnostic testing in a general maximum likelihood framework. Under certain conditions the diagnostic and the sensitivity are asymptotically independent, therefore, the sensitivity analysis is important. Chapter 3 studies the sensitivity of random effects estimators in the one-way error component regression model. It suggests an explanation of the simulation evidence of Maddala and Mount (1973) that the properties of the feasible GLS estimator are not affected by the choice of the first-step estimator used for the variance matrix. Combination of forecasts from different levels is the second part of this thesis. As often happens in practice, data are available at different levels or at different frequencies. In Chapter 4 a two-level hierarchical model is proposed, which allow us to update the micro forecast on the basis of the macro forecast. The proposed method has a natural interpretation and shows good performance in Monte Carlo simulations. Chapter 5 extends the method to time series and gives an empirical application for the European zero rates, making optimal use of the combination of quarterly and monthly observations.
|Qualification||Doctor of Philosophy|
|Award date||12 Dec 2006|
|Place of Publication||Tilburg|
|Publication status||Published - 2006|