@misc{oai:ir.soken.ac.jp:00000774,
 author = {王, 健歡 and オウ, ケンカン and WONG, Kin Foon Kevin},
 month = {2016-02-17, 2016-02-17},
 note = {This thesis summarizes statistical analysis of some multivariate heteroscedastic<br /> time series data, including 2 sets of data from physiological experiments and<br /> 2 sets of EEG data about anaesthesia and coma.<br />　　The aim of this thesis is to provide a statistical tool for analyzing multi-<br />variate data which contains non-stationary and heteroscedastic characteristics.<br />　　The main contribution of this thesis is that we combine the linear state<br /> space model and GARCH model to develop a state space-GARCH model.<br />The state space-GARCH model can describe the non-stationary characteristics<br /> of the system noise variance. In particular we adopt a special structure of<br /> the linear state space model to decompose a data into components by their<br /> frequencies. Combining a heteroscedasticity model and a state space model<br /> is carried out by fully utilizing the information of innovations and expected<br /> values from the filtering process.<br />　　Another contribution of the thesis is that we extend Akaike's NCR from<br /> constant noise variance to heterogeneous noise variance in order to study time-<br />varying causality. By applying heteroscedasticity models, the phenomenon of<br /> an evolving causality relationship can be depicted.<br />　　All these methods are illustrated by their application to EEG data including<br /> the study of consciousness under anaesthesia and coma, and also to a physical<br /> data of head and finger movement., 総研大甲第948号},
 title = {Multivariate Times Series Analysis of Heteroscedastic Date, with Application to Neuroscience},
 year = {}
}