Tid: 16 januari 2006 kl 1515-1600
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Johan Kilander
Titel: Dimension reduction techniques and multivariate GARCH modeling. (Examensarbete)
Sammanfattning: The purpose of a multivariate GARCH (MGARCH) model is to make the covariance matrix of a set of risk factors conditional on past events. When the number of risk factors is large, accurate and robust estimation of most MGARCH models will not be feasible. The objective of this thesis is to construct an MGARCH model that efficiently overcomes the estimation problems but still provides a great deal of flexibility. To this end two existing MGARCH models are considered: the Orthogonal (O) GARCH model and the Constant Conditional Correlation (CCC) GARCH model. The strengths of these models are then built upon to form a new MGARCH model referred to as the CCC-OGARCH. The CCC-OGARCH imposes constant conditional correlations on a reduced set of risk factors. To test the validity of this assumption a model with state-dependent correlations is proposed. Statistical testing and Value-at-Risk computations indicate that the risk factors could be appropriately modeled in a CCC-OGARCH framework.