DSpace - Tor Vergata >
Facoltà di Economia >
Tesi di dottorato in economia >

Please use this identifier to cite or link to this item: http://hdl.handle.net/2108/1409

Full metadata record

DC FieldValueLanguage
contributor.advisorProietti, Tommaso-
contributor.authorFenga, Livio-
date.accessioned2010-08-06T09:24:09Z-
date.available2010-08-06T09:24:09Z-
date.issued2010-08-06T09:24:09Z-
identifier.urihttp://hdl.handle.net/2108/1409-
description20. cicloen
description.abstractPresupposto fondamentale per lo studio di molti fenomeni dinamici, ad esempio nel campo dell’analisi dei segnali, dell’identificazione di sistemi e delle serie storiche, e’ la modellazione del processo aleatorio sottostante. Nel caso lineare, il problema e’ affrontato con modelli della classe ARMA (autoregressivi a media mobile). Tuttavia, e’ probabile che serie storiche realizzazione di fenomeni reali contengano pronunciate dinamiche non lineari, come ad esempio comportamenti dipendenti da regimi e chaos. Tali caratteristiche, invece, possono venire adeguatamente catturate da modelli appartenenti alla classe SETAR. In questo lavoro, il problema della scelta dell’ordine “corretto” per queste tipologie di modelli e’ affrontato nell’ambito della teoria bootstrap associata a selettori di ordine di largo impiego. In particolare, verra’ presentato un approccio bootstrap per l’identificazione dell’ordine, le cui performances saranno valutate con tecniche di simulazione Monte Carlo.en
description.abstractModeling the underlying stochastic process is one of the main goals in the study of many dynamic phenomena, such as signal processing, system identifcation and time series. In the linear case, the issue is often addressed within the framework of ARMA (Auto-Regressive Moving Average) paradigm. However, many real-life time dependent processes, are likely to show pronounced nonlinear dynamics, like regime dependent behaviors and chaos, that can be adequately captured by models of the class SETAR (Self Exciting Threshold Autoregressive Models). In this work, we bring the problem of the choice of the "correct" order for these type of models in the framework of bootstrap theory in conjunction with widely employed order selectors. In particular, a bootstrap-based order identification method will be presented and its performances assessed via Monte Carlo simulations.en
description.tableofcontentsPART I Bootstrap based ARMA order identification 1. Introduction 2. Akaike information criterion for the selection of ARMA models 3. Bootstrap for dependent data 3.1 The employed scheme: the sieve bootstrap 3.2 Definition of AIC* and the proposed B-MAICE method 4. On the asymptotic distribution of the orders selected via AIC and AIC* for ARMA processes 4.1 Two important results of Shibata and Hannan 4.2 On the asymptotic equivalence of the distribution of the orders chosen by both AIC and AIC* for ARMA models 5. The B-MAICE method and empirical results 5.1 The procedure 5.2 Some remarks on the method 5.3 Results and discussion 5.3.1 Empirical evidences 5.3.2 Conclusions. - References. - PART II Bootstrap based order determination for ARMA models: a comparison between different order selection criteria 1. Introduction 2. Order selection for time series models 2.1 The employed identification criteria 2.2 Arma model selection through minimization of selection criteria 3. The bootstrap method 3.1 The bootstrapped selection criteria 3.2 The applied bootstrap scheme 3.3 The presented B-MSCE procedure 4. Empirical Study 4.1 The experiments 4.2 Results 4.3 Final remarks - References. - PART III Bootstrap based order selection for SETAR models 1. Introduction 2. Akaike information criterion for SETAR models 3. Estimation and selection strategy for l-regimes SETAR models 4. The employed bootstrap scheme 5. Definition of beta-AIC and the proposed beta-MAICE procedure 6. Simulation study 6.1 Experimental design and simulations 7. Artificial neural network approach for the assessment of the robustness of the proposed method 7.1 Choice of the models belonging to the subset a 7.2 The employed selection criterion for artificial neural network 7.3 Empirical results - References - Conclusionsen
format.extent726236 bytes-
format.mimetypeapplication/pdf-
language.isoenen
subjectorder selectionen
subjectinformation criteriaen
subjectARMA modelsen
subjectSETAR modelsen
subjectsieve bootstrapen
subjectstationary bootstrapen
subject.classificationSECS-P/05 Econometriaen
titleTime series bootstrap-based model order selectionen
typeDoctoral thesisen
degree.nameEconometria ed economia empiricaen
degree.levelDottoratoen
degree.disciplineFacoltà di economiaen
degree.grantorUniversità degli studi di Roma Tor Vergataen
date.dateofdefenseA.A. 2009/2010en
Appears in Collections:Tesi di dottorato in economia

Files in This Item:

File Description SizeFormat
PhD Thesis.pdf709KbAdobe PDFView/Open

Show simple item record

All items in DSpace are protected by copyright, with all rights reserved.