DSpace Repository

New Development on Market MicroStructure and MacroStructure: Patterns of U.S. High Frequency Data and A Unified Factor Model Framework

Show simple item record

dc.contributor.advisor Mullhaupt, Andrew Peter en_US
dc.contributor.author Dong, Xu en_US
dc.contributor.other Department of Applied Mathematics and Statistics en_US
dc.date.accessioned 2017-09-20T16:52:57Z
dc.date.available 2017-09-20T16:52:57Z
dc.date.issued 2013-12-01 en_US
dc.identifier.uri http://hdl.handle.net/11401/77588 en_US
dc.description 100 pgs en_US
dc.description.abstract In this thesis we study the problem of modeling cross-sectional volatility structure of U.S. stock market. The thesis contains two parts. In the first part we identify a particular volatility spike phenomenon, that the cross-sectional volatility is significantly stronger at every 5 minute. An empirical spike study is conducted on individual stock level by applying Lee-Mykland jump detector. By constructing spike ratio statistics, the evolving paths of this spike phenomenon is studied during the 1993 to 2009 time period. In the second part, we model the volatility structure using factor structures. Particular, we propose two approaches to build a better model for small sample size problem, which is a common issue for high dimension models. We first study the fisher information matrix and derive Jeffrey's prior for the factor model. We extend the EM algorithm method for factor parameter estimation by applying the Jeffrey's prior, which turns to be more robust in the sense that risk would not be underestimated under small sample size. Next we extend the statistical factor model to be able to process empirical information.The fundamental(empirical) factor model and the statistical(hidden) factor model are widely used in factor analysis. The fundamental factor model is empirical, biased, and has small estimation variance. On the other hand, the statistical factor model is subjective, unbiased, but has a bigger estimation variance. In this paper, a new factor model called EH Factor Model (Empirical-Factor--and--Hidden-Factor Factor Model) is introduced to combine the two models under a unified framework. The EH model allows to include exogenous information to help reducing the number of parameters, and meanwhile it maintains a statistical factor structure so that the idiosyncratic variance is kept. In this way, it reduces the estimation variance and the bias. We also compare the EH model with a widely used hybrid approach, which firstly apply the fundamental factor model and then treat the residuals with the statistical factor model. Compare with this approach, The EH model does not require to assume that the column space of statistical factor exposures is orthogonal to the fundamental factor exposures, and is also shown to be more robust in a sense of less estimation variance and bias. en_US
dc.description.sponsorship This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree. en_US
dc.format Monograph en_US
dc.format.medium Electronic Resource en_US
dc.language.iso en_US en_US
dc.publisher The Graduate School, Stony Brook University: Stony Brook, NY. en_US
dc.subject.lcsh Applied mathematics en_US
dc.subject.other cross-sectional volatility, EH model, EM algorithm, factor model, Jeffrey's Prior, spike phenomenon en_US
dc.title New Development on Market MicroStructure and MacroStructure: Patterns of U.S. High Frequency Data and A Unified Factor Model Framework en_US
dc.type Dissertation en_US
dc.mimetype Application/PDF en_US
dc.contributor.committeemember Rachev, Svetlozar en_US
dc.contributor.committeemember Xing, Haipeng en_US
dc.contributor.committeemember Smith, Noah en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account