"Modeling, forecasting and diagnostic testing for financial and econometric time series"

拟议的研究是在商业、经济和金融的时间序列分析和统计领域。它的主要目标是：(I)开发高频数据波动性的新模型；(Ii)设计有效的波动性测量和预测方法；(Iii)构建诊断测试，以检测市场动态的变化。这项研究的背景是当今的金融市场，交易和价格变化以毫秒为单位进行记录。统计学家和计量经济学家在20世纪80年代开发的经典量化模型无法解释这种高频数据的特征。自20世纪90年代末以来，已经引入了更合适的模型，然而，仍然缺乏解释日内和日间市场行为的全面模型。拟议的研究旨在缩小这一差距，并导致更好地理解市场动态如何在不同的时间尺度上表现出来。它的主要重点是研究波动率，波动率衡量的是大宗商品价格随时间的变化，在风险管理中具有重要应用。特别关注的将是预测波动性的有效方法。另一条研究主线是开发交易事件之间持续时间的模型。经验研究表明，持续时间之间存在很强的相关性，因此需要能够在数据中捕捉这种长期相关性的模型。最后，这项拟议的研究旨在考察市场动态受一组潜在变量支配的模型。在这种背景下，高度相关的问题是如何检测潜在结构中的制度变化，以及在多大程度上可以预测市场动态。拟议的研究将产生一些方法，帮助加拿大金融业的研究人员和从业者更好地了解金融市场的动态，获得更准确的预测，并更好地评估风险与回报之间的权衡。它包括培训5名本科生、8名硕士和1名博士生，他们将获得扎实的金融时间序列分析知识，强大的编程和解决问题的技能，并获得前沿跨学科研究的经验。

The proposed research is in the area of Time Series Analysis and Statistics in Business, Economics and Finance. Its main objectives are (i) developing new models for the volatility of high frequency data; (ii) designing effective methods for measuring and forecasting volatility; (iii) constructing diagnostic tests for the detection of changes in the market dynamics. The context of this research is today's financial markets where transactions and price changes are recorded on the millisecond time scale. The classical quantitative models developed by statisticians and econometricians in the 1980s fail to explain characteristic features of such high frequency data. More suitable models have been introduced since the late 1990s, however, there is still a lack of comprehensive models explaining market behaviour both on the intra- and inter-daily level. The proposed research aims to close this gap and lead to a better understanding of how market dynamics manifests itself on different time scales. Its main focus is on the study of volatility which measures the variation of commodity prices over time and has important application in risk management. A special focus will be on efficient methods for forecasting volatility. Another main line of research is the development of models for the durations between transaction events. Empirical studies have shown strong correlations among the durations, so there is a need for models which can capture this long-range dependence in the data. Finally, the proposed research aims to look at models in which the market dynamics is governed by a set of latent variables. Highly relevant questions in this context are how regime switches in the latent structure can be detected, and to what extent it is possible to forecast market dynamics. The proposed research will yield methods which help researchers and practitioners in the Canadian financial industry to better understand the dynamics of financial markets, obtain more accurate predictions and better assess risk-return tradeoffs. It involves the training of five undergraduate, eight Master's and one PhD students who will acquire solid knowledge in financial time series analysis, strong programming and problem solving skills, and gain experience in frontier interdisciplinary research.