Back to search

Aristotle University Of Thessaloniki (AUTH)

Αιτιότητα κατά Granger και ανάλυση πολυ-μεταβλητών χρονοσειρών με εφαρμογή στη νευροφυσιολογία

Abstract

dc:description

The aim of the present thesis is the valuation and also the development of new methods, regarding the research field of connectivity analysis of complex dynamical systems. Also, this thesis is strongly multidisciplinary, given that we use the methods that we developed in order to shed light to problems of the field of neurophysiology and finance. After extensive literature review, we used a large number of causality measures, popular in the research of multi-variate time series analysis. We also used dynamical systems from different categories, such as non-linear difference equations, delay differential equations, a system that simulates the electroencephalogram and the stochastic process of vector autoregressive model in many scenarios regarding the dynamical regime, the number of variables and the coupling strength. The main conclusion of this study is that the dimension reduction measures are more efficient in the estimation of the true coupling structure. The causality network characteristics provide useful information for the topology of the causal effects among the system's variables. We evaluate the causality network characteristics in terms of their ability to classify multi-variate time series in different coupling structure categories. For this we use different setups in terms of the dynamical regime and coupling strength of the system under study. The conclusion is that network characteristics related to the strength distribution and the clustering of the nodes hold the crucial discriminative information. The estimation of causality effects is a hard task when important variables of the system are not observed. In this case non-existing causal effects are estimated as existing. Also, existing causal effects are not estimated as such when the impact of the hidden variable is strong, because of the common input that introduces correlations to the system. We developed a method for the estimation of causality and instantaneous causality from multi-variate time series, that shows great performance. We also showed through an analytical example, that the additional causal effects that arise when latent confounders are present are not necessarily false. Concerning instantaneous causality, we applied the proposed method on electroencephalogram recordings with epileptiform discharges. The results showed that hidden variables are present during the electroencephalogram recordings, the impact of which increases during the epileptiform discharges. In the electroencephalogram applications we used uni-variate and multi-variate means to analyse the multi-variate time series. The results showed that the duration of an induced (with transcranial magnetic stimulation) epileptiform discharge is, to some extent, correlated with the dynamical regime of certain brain regions in the early stage of the epileptiform discharge. Also, the connectivity analysis showed that the brain network undergoes critical changes before, during and after the epileptiform discharge. More precisely, there exist three discrete stages during the epileptiform discharge. Using the same recordings we applied uni-variate methods of time series analysis and we inferred that the epileptic brain has many covert states of excitability during the interictal period.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2017

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Koutlis, Christos
  • Κουτλής, Χρήστος

Subjects

dc:subject × 14

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/40062

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Koutlis, Christos; Κουτλής, Χρήστος. Αιτιότητα κατά Granger και ανάλυση πολυ-μεταβλητών χρονοσειρών με εφαρμογή στη νευροφυσιολογία. Aristotle University Of Thessaloniki (AUTH), 2017. http://hdl.handle.net/10442/hedi/40062