Dissertation
THREE ESSAYS ON CONTINUOUS AND DISCRETE TIME SERIES MODELS
Doctor of Philosophy (PhD), Washington State University
01/2015
Handle:
https://hdl.handle.net/2376/6252
Abstract
Time series models have been widely used in finance and economics. In the first essay, an extreme value test is considered to evaluate a testing predictive ability. The exact asymptotic distribution function of the test under nuisance parameters is attained. The test is designed to evaluate comparisons of a set of models against a benchmark model when a finite number of comparisons is deliberated. To surpass nuisance parameters, consistent estimators are exploited. The test captures the effect of estimation uncertainty on the relative forecast performance, it allows the forecasts to be produced by a general estimation method and it is easy to compute. The assumptions are realistic and mild in nature. Illustrative examples are then employed and the proposed methodology is applied to decide, which in a given set of models provides the most accurate predictive ability.
The second essay investigates how the misspecification in the marginal and copulas may affect the correlation forecasting accuracy of the Copula-DCC and Copula-GAS models. Different specifications of volatility models are also tested on the correlation forecasting ability. These models are fitted to both simulated data and real data. In real data, the realized correlation is taken as a proxy for actual correlation and is used for computing the loss function.
In the third essay, a convenient Laplace transform of the bivariate supremum and the time last time a process is attained its supremum is provided. This is demonstrated when the underline process is a subordinate linear Brownian motion with negative drift. To illustrate the effectiveness of this formulation, a Lévy measure of an exponential subordinator is exploited. In this case, in addition to their Laplace transforms, both the probability density of the supremum and the time last time its supremum is attained are explicitly determined.
Metrics
10 File views/ downloads
42 Record Views
Details
- Title
- THREE ESSAYS ON CONTINUOUS AND DISCRETE TIME SERIES MODELS
- Creators
- Jun Wang
- Contributors
- Stergios B Fotopoulos (Advisor)Sung K. Ahn (Committee Member)Charles L. Munson (Committee Member)George Jiang (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Carson College of Business
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Number of pages
- 131
- Identifiers
- 99900581837201842
- Language
- English
- Resource Type
- Dissertation