http://www.netyi.net/Category/80论文下载基地_最新资料_得益网得益网 http://www.netyi.net/training/2d9446db-f9c8-4d8c-baac-4cea779eccc3论 文 题　目：基于数据挖掘的指纹识别系统的研究与设计<br/>学 科 名　称：软件工程<br/><br/>目 录<br/><br/>文摘<br/>英文文摘<br/>西北工业大学学位论文知识产权声明书及原创性声明<br/>第1章绪论<br/>1.1生物识别技术介绍<br/>1.2指纹识别技术<br/>1.2.1背景介绍<br/>1.2.2指纹识别中的基本概念<br/>1.2.3自动指纹识别系统构成<br/>1.2.4指纹识别系统的性能评价<br/>1.2.5国内外研究现状<br/>1.3论文的主要工作及结构安排<br/><br/>第2章指纹图像预处理的研究与应用<br/>2.1预处理概述<br/>2.1.1概述<br/>2.1.2指纹质量评估<br/>2.2图像的归一化<br/>2.3指纹图像增强<br/>2.3.1方向场的计算<br/>2.3.2 Gabor滤波增强<br/>2.4指纹图像梯度锐化<br/>2.4.1梯度锐化方法<br/>2.5指纹图像分割<br/>2.6指纹图像二值化<br/>2.6.1最大类间方差法<br/>2.6.2自适应二值化法<br/>2.7指纹图像的细化<br/>2.7.1指纹图像细化后的处理<br/>2.8小结<br/><br/>第3章指纹图像的特征提取的研究与应用<br/>3.1指纹细节点提取概述<br/>3.2常规指纹局部细节特征提取方法介绍<br/>3.2.1指纹的纹理特征<br/>3.2.2结构化的指纹特征描述方法<br/>3.2.3将细节点特征和脊线采样相结合的指纹特征表示<br/>3.2.4图像几何特征<br/>3.2.5图像的统计特征<br/>3.2.6图像的变换系数特征<br/>3.2.7图像的代数特征<br/>3.2.8利用神经网络提取图像特征<br/>3.3常规指纹局部细节特征提取方法<br/>3.4伪特征结构的滤除<br/>3.4.1指纹的伪特征结构<br/>3.4.2伪特征的滤除算法<br/>3.5小结<br/><br/>第4章基于数据挖掘的指纹识别系统的设计<br/>4.1指纹图像的分类<br/>4.2指纹样本的类别判断<br/>4.2.1属性的相关度分析<br/>4.2.2K-最临近分类法<br/>4.3数据仓库与数据挖掘技术介绍<br/>4.4基于指纹图像的数据挖掘<br/>4.4.1指纹数据仓库的建立<br/>4.4.2指纹数据仓库数据模型的建立<br/>4.4.2基于指纹特征检索的主要特点<br/>4.5基于数据挖掘的指纹识别系统的总体结构设计<br/>4.5.1指纹样本数据采集<br/>4.5.2指纹数据预处理<br/>4.5.3指纹图像特征的提取<br/>4.5.4指纹图像数据特征立方体<br/>4.5.5指纹图像数据挖掘模块<br/>4.5.6基于神经网络的指纹识别模块<br/>4.5.7实验结果及结论<br/>4.6小结<br/><br/>结束语<br/>参考文献<br/>硕士期间发表论文<br/>致谢<br/><br/>摘要<br/> 指纹 的 唯 一性和不变性，使指纹识别技术成为当今最广泛的身份认证和识别<br/>技术之一。目前，指纹识别技术己被广泛地应用于公安、海关、银行、网络安全<br/>等领域，具有较高的理论意义和实用价值。<br/> 在指 纹 自 动识别过程中，存在指纹图像噪声的影响，需要对指纹信息进行去<br/>噪、增强处理并且需要进行大量的指纹匹配处理。这种情况下，指纹的识别率不<br/>是很高，而且运算速度很慢。本文深入研究了指纹图像预处理的处理过程，给出<br/>了一种数据挖掘技术与模式识别技术相结合的指纹图像识别方法。为解决指纹自<br/>动识别技术中存在的种种困难找到切实可行的新途径。<br/> 本文 主 要 对指纹的预处理、特征提取和基于数据挖掘的指纹识别方法等进行<br/>深入研究。主要研究内容如下:<br/> 1、 在 指 纹图像预处理环节<br/>采用 了 基 于Gabor滤波增强的算法，该方法增强效果良好，计算量小;采用<br/>局部阀值的方法对指纹进行二值化，再对指纹进行细化，得到清晰的指纹点线图。<br/> 2 、 对 指 纹的特征提取算法进行了研究<br/>采 用的 求 取8邻域纹线跟踪的算法能够准确地提取出指纹图像的细节特征，<br/>并表现出较强的抗干扰性。再利用脊线跟踪的后处理算法去除指纹的伪细节特<br/>征，如毛刺、小桥和小孔等。采用这种方法对细节特征进行后处理，能有效地去<br/>除伪特征点，为提高指纹匹配的速度和准确性奠定了良好的基础。<br/> 3、 对 指 纹图像进行数据挖掘的方案进行了研究<br/> 在 研 究 数据挖掘模型的基础上，在指纹图像数据挖掘体系方面，提出了基于<br/>数据仓库的指纹图像数据挖掘流程，在此基础上提出一个指纹图像数据挖掘集成<br/>框架，以实现指纹图像数据挖掘过程中所涉及到的数据准备、建立数据仓库、指<br/>纹图像数据挖掘、运用神经网络进行模式识别等技术环节。<br/><br/>关键词:指纹识别系统，指纹图像预处理，特征提取、数据挖掘，数据仓库2008-10-13 20:05:42 http://www.netyi.net/training/312a70c4-8569-41dd-aa99-d73e95eb0044论 文 题　目：基于Web+Services的实时数据访问技术的研究<br/>学 科 名　称：计算机应用技术<br/><br/>目 录<br/><br/>文摘<br/>英文文摘<br/>声明<br/>第1章引言<br/>1.1选题背景及意义<br/>1.3国内外研究状况<br/>1.3.1国外研究现状<br/>1.3.2国内研究现状<br/>1.4本文主要工作及构成<br/>1.5创新点<br/><br/>第2章Web Services技术研究<br/>2.1 Web Services概述<br/>2.1.1 Web Services的定义及特点<br/>2.1.2 Web Services的体系结构<br/>2.2 Web Services相关技术<br/>2.2.1 Web Services技术的基石：XML<br/>2.2.2 Web Services的消息传递方式：SOAP<br/>2.2.3 Web Services的描述方式：WSDL<br/>2.2.4 Web Services的注册发布：UDDI<br/>2.3基于Web Services实现实时数据服务的优势<br/>2.7本章小结<br/><br/>第3章Web Services的实时性能分析<br/>3.1问题的提出<br/>3.1.1 Web Services架构的优势<br/>3.1.2项目的需求<br/>3.1.3存在的问题<br/>3.2 Web服务性能分析<br/>3.2.1 Web服务应用模式<br/>3.2.2 Web服务性能的衡量指标<br/>3.2.3网络传输因素<br/>3.2.4 XML处理因素<br/>3.2.5 SOAP协议因素<br/>3.2.6 Web服务运行环境因素<br/>3.3本章小结<br/><br/>第4章提高Web Services实时性策略<br/>4.1 XML压缩技术<br/>4.1.1数据压缩技术<br/>4.1.2模式专用压缩<br/>4.1.3为XML专门设计压缩算法<br/>4.1.4压缩的评估公式<br/>4.2压缩SOAP消息<br/>4.2.1压缩SOAP Message的方法<br/>4.2.2压缩方法的实现<br/>4.3利用VTD-XML提高Web服务的性能<br/>4.3.1 VTD-XML简介<br/>4.3.2与DOM、SAX解析器的性能比较<br/>4.3.3基于VTD-XML的解析模板<br/>4.3.4 Web服务响应模板模式<br/>4.4其它缓存优化策略<br/>4.4.1客户端缓存<br/>4.4.2多级缓存<br/>4.5本章小结<br/><br/>第5章基于Web Services实时数据访问技术的实现<br/>5.1总体设计架构<br/>5.2基于事件的实时数据发布机制<br/>5.3监控数据的时效性<br/>5.3.1数据时效性的基本概念<br/>5.3.2实时数据服务系统中数据时效性的设计<br/>5.4实时数据服务的实现<br/>5.5服务调用测试<br/>5.5.1 wsCaller工具调用测试<br/>5.5.2测试结果分析<br/>5.5.3远程Web服务的调用<br/>5.6本章小结<br/><br/>第6章结束语<br/>6.1工作总结<br/>6.2工作展望<br/><br/>参考文献<br/>致谢<br/>作者在攻读硕士学位期间发表的学术论文<br/><br/><br/> 摘 要<br/> 随着 互 联 网软件技术及其应用迅速发展，基于WebS ervices的分<br/>布式计算模式日益成为软件技术和应用发展的趋势，Web Services为<br/>分布式计算提供了一种新的范例。WebServices技术是一种面向开放互<br/>联网协议的软件应用，它通过XML消息及协议完成与其它应用软件的<br/>交互，实现更大范围内系统间的互联、互通和互操作。<br/> 在企 业 的 信息化发展过程中，信息集成是企业发展的总趋势，如<br/>何把工业企业生产现场中各种异构环境下的实时数据集成起来，形成<br/>统一的实时数据库，实时地通过局域网和广域网传送到全企业和上级<br/>主管部门，达到真正的网络化管理，是目前中小型企业面临的问题。<br/>多数的企业信息集成方案都是采用传统的中间件技术，集成双方之间<br/>实现的是紧藕合机制，导致系统的柔软性、互联性和可扩充性受到限<br/>制。而Web Services技术作为一种新型的分布式对象技术，具有完好<br/>封装、松散藕合和高度可集成能力等特点，能够很好地满足企业成员<br/>的动态性要求，井解决成员之间应用集成的架构相异问题，从而使企<br/>业的应用集成环境具有良好的可扩展性和易维护性。基于Web<br/>Services的实时数据访问技术，采用面向服务架构(SOA)的软件设计<br/>方法，能够把使来自工业现场的实时数据信息发布为Web服务，同时<br/>也为企业的〔RP, CRM, PDM等系统提供实时数据访问服务，服务请求<br/>者可以在B/S模式下访问实时数据库。<br/> 目前 ， 基 于WebS ervices架构下系统服务响应的实时性还不能完<br/>全满足企业信息集成的要求，本文在分析Web Services性能现状的基<br/>础上，着重从优化Web Services性能方面入手来研究如何提高服务调<br/>用的实时性:影响Web Services的时间响应速度的因素主要有三个:<br/>网络传输时间、消息处理时间和服务执行时间。在网络传输阶段，通<br/>过压缩XML文件来缩短网络传输的时间，提出了压缩SOAP消息和模式<br/>压缩的方法;在消息处理阶段中，XML解析、反序列化、序列化是三个<br/>最消耗时间的阶段，是Web Services性能的瓶颈，尤其是在有效负荷<br/>增大的时候。要提高Web Services的性能，优化XML解析、序列化与<br/>反序列化过程是关键。本文引入VTD-XML解析器，实验分析了XML解<br/>析器DOM, SAX和VTD-XML各自的特点，在速度和性能方面VTD-XML能<br/>较好满足实时性的要求，根据工业现场实时数据访问的特点和VTD-XML<br/>解析器的工作原理，设计了基于VTD-XML的解析模板缓存来加快XML<br/>文档的解析速度，用MD5算法为每个SOAP消息请求生成一个唯一的解<br/>析模板ID，并用VTD-XML提供的API操作进行模板缓存的管理，实验<br/>证明了VTD-XML解析模板缓存能进一步提高Web Services的性能，满<br/>足实时数据访问服务的要求。最后在J2EE平台下使用日EA公司的<br/>Weblogic实现7基于Web Services的实时数据访问服务功能，通过在<br/>系统中配置自己开发的基于VTD-XML解析模板缓存，使得该实时数据<br/>访问服务能能够满足用户、计算机及企业的其它系统如〔RP, PDM, CRM<br/>等实时性的要求。<br/><br/>关键词:Web Services 实时性信息集成性能优化2008-10-13 19:59:45 http://www.netyi.net/training/71498e23-ee9d-4d41-aa67-acf1967f9e7a论 文 题　目：基于J2EE平台在线银行系统的实现<br/>学 科 名　称：计算机应用技术<br/><br/>目 录<br/><br/>文摘<br/>英文文摘<br/>声明<br/>第1章绪论<br/>1.1课题的提出及实际意义<br/>1.1.1国外在线银行的发展情况<br/>1.1.2国内在线银行的发展现状<br/>1.2开发技术应用概述<br/>1.3课题研究的目的和意义<br/>1.4课题的来源和研究内容<br/><br/>第2章基于J2EE平台开发技术体系结构<br/>2.1 J2EE与.NET平台比较<br/>2.1.1技术概述：<br/>2.1.2技术比较：<br/>2.1.3整体评价：<br/>2.2 J2EE概述及其优势<br/>2.2.1 J2EE概述<br/>2.2.2 J2EE技术优势<br/>2.2.3 J2EE平台体系简介<br/>2.2.4 J2EE的层次<br/>2.2.5 EJB组件简介<br/>2.3 JSP技术<br/>2.4 Struts的结构和流程简介<br/>2.4.1 MVC设计模式<br/>2.4.2 Struts框架<br/>2.4.3基于MVC模式的Struts框架的开发步骤<br/>2.5 Hibernate简介<br/>2.6本章小结<br/><br/>第3章在线银行应用系统的实现<br/>3.1在线银行系统设计<br/>3.1.1基于J2EE应用程序的开发步骤<br/>3.1.2系统的总体结构设计<br/>3.1.3系统功能模块设计<br/>3.1.4系统的体系结构<br/>3.1.5 MVC结构<br/>3.1.6数据库的结构分析与设计<br/>3.1.7系统开发环境<br/>3.2业务逻辑层EJB的实现<br/>3.2.1实体Bean的设计与实现<br/>3.2.2会话Bean的设计与实现<br/>3.2.3业务逻辑其它package功能<br/>3.2.4表示层的开发及实现<br/>3.2.5 Hibernate使用<br/>3.2.6国际化多语言显示<br/>3.2.7系统的配置运行<br/>3.3本章小结<br/><br/>第4章基于安全机制在线银行构架及应用程序实现<br/>4.1基于安全机制在线银行基本构架<br/>4.2基于J2EE技术在线银行安全应用实现<br/>4.2.1应用程序安全<br/>4.2.2在线应用程序安全<br/>4.2.3应用程序Web安全管理<br/>4.2.4 EJB安全<br/>4.3 Weblogic安全机制<br/>4.4本章小结<br/><br/>结论<br/>附录<br/>参考文献<br/>攻读学位期间发表的学术论文<br/>致谢<br/><br/>摘要<br/> 近 年 来 ，随着电子商务的迅速发展，在线银行也得到世界各国的普遍关<br/>注，它利用Internet, Intranet及相关技术处理传统的银行业务及支持电子商<br/>务的在线支付。它的出现改变了传统的银行交易形式，引发了一场新的经济<br/>革命。由于在线银行内部业务逻辑复杂，安全性要求高，商务形式发展变化<br/>快，这要求改变传统的Web技术设计模式，以适应当前在线银行发展技术<br/>的需求。<br/> 本 课题 利 用J2EE的体系结构结合Struts的MVC架构和Hibernate技术<br/>开发在线银行系统，同时兼顾在线银行的安全体系开发技术，并且基于<br/>J2EE在线银行系统的设计进行了项目实践。系统开发利用了J2EE企业开发<br/>架构模式，较好地利用XML跨平台机制与原有ATM和POS旧系统实现数<br/>据跨平台交换;同时利用J2EE 和Hibernate构架实现新旧数据库系统数据<br/>互访，用户端利用J2EE新技术展现给用户一个全新界面，而系统底层则有<br/>多个应用系统， 多个数据库并存的局面，可避免应用新技术应用初期运行不<br/>稳定带来的潜在风险;在国际化方面采用Struts结构来实现多种语言页面的<br/>显示;在系统安全上利用Weblogic安全认证开发模式实现在线银行安全机<br/>制。<br/> 本 文基 于 在线银行系统的开发，分析了在线银行的特点和具体需求，深<br/>入地研究了J2EE 架构及所包含的各种技术， 提出了J2EE, Struts的<br/>MVC, XML和Hibernate构架，对在线银行应用系统的设计与实现，详细<br/>介绍了开发的具体步骤和细节。<br/><br/>关键词在线银行;组件:J2EE; EJB; MVC2008-10-13 19:55:27 http://www.netyi.net/training/bbe9b8f7-2614-4b8e-95f2-375e49e9d829论 文 题　目：残缺足迹图像拼接技术的研究与实现<br/>学 科 名　称：计算机应用技术<br/><br/>目 录<br/><br/>文摘<br/>英文文摘<br/>独创性说明和关于论文使用授权的说明<br/>引 言<br/>1文献综述<br/><br/>1.1痕迹、足迹和足迹特征<br/>1.2数字图像处理<br/>1.3透视投影变换及其反变换<br/>1.4计算几何应用<br/><br/>1.4.1多项式插值<br/>1.4.2样条曲线的基本概念<br/>1.4.3 B样条的计算<br/>1.4.4 B样条曲线在形状描述中的应用<br/><br/>1.5图像的基本属性<br/><br/>1.5.1分辨率<br/>1.5.2像素深度<br/>1.5.3真彩色.伪彩色<br/><br/>1.6图像格式<br/>1.7 Microsoft Foundation Class<br/>1.8 VC++ 6.0 I DE(集成开发环境)<br/><br/>2残缺图像拼接技术<br/><br/>2.1可行性分析和需求分析<br/><br/>2.1.1需求陈述<br/>2.1.2可行性分析<br/>2.1.3系统硬件要求<br/>2.1.4系统开发的软件要求<br/><br/>2.2算法基础<br/><br/>2.2.1透视投影处理算法<br/>2.2.2痕迹边缘的点取和轮廓拟和算法<br/>2.2.3图像的特征匹配算法<br/>2.2.4位图缩放算法<br/>2.2.5位图旋转算法<br/>2.2.6图像整合和缝合处理算法<br/><br/>2.3残缺图像拼接系统的设计<br/><br/>2.3.1残缺图像拼接系统的总体设计<br/>2.3.2位图读入与显示功能设计<br/>2.3.3图层的生成功能设计<br/>2.3.4图层的激活功能设计<br/>2.3.5图层的移动功能设计<br/>2.3.6图层的融合功能设计<br/><br/>2.4系统实现<br/><br/>2.4.1开发环境的选择<br/>2.4.2程序结构和数据结构<br/><br/>2.5验证<br/><br/>2.5.1系统使用说明<br/>2.5.2试验准备<br/>2.5.3系统功能检验<br/>2.5.4检验结果分析<br/><br/>结论<br/>参考文献<br/>在学研究成果<br/>致 谢<br/><br/><br/>摘要<br/> 足迹 是 存 储信息的一种载体，它反映了与人足有关的各个形态的形象，公安人员能<br/>通过它对案件足迹与嫌疑人足迹是否同一做出认定，为侦察机关和审判机关及时准确的<br/>提出破案线索和诉讼证据。<br/> 由北 京 科 技大学计算机系王秀美教授与北京刑事科学研究所合作开发的计算机辅助<br/>足迹检验系统已基本实现自动化足迹检验。该系统利用犯罪现场采集到的犯罪嫌疑人足<br/>迹信息，通过计算和分析，提取犯罪嫌疑人的身体特征，以便于公安机关进行刑事侦<br/>查。犯罪现场的情况往往比较复杂，大多数现场足迹为不完整的或部分不清晰的足迹，<br/>清晰、完整的犯罪嫌疑人足迹图像较难得到，这就限制了计算机辅助足迹检验系统的应<br/>用。<br/> 大量 的 残 缺足迹图像是宝贵的破案线索和诉讼依据，使用人工方式很难完成拼接工<br/>作，即便拼接成功也存在费工费时、可调整性差、误差大等问题，这样就浪费了这些犯<br/>罪现场的残缺图像。因此，利用计算机技术构建残缺图像拼接子系统，可以通过计算机<br/>技术手段将残缺足迹图像拼接成完整的足迹图像，以供计算机辅助足迹检验系统和其他<br/>痕迹分析系统使用。<br/> 残缺 足 迹 图像拼接技术的研究主要涉及痕迹学、数字图像处理、计算机图形学、面<br/>向对象程序设计、软件工程等理论和技术。算法包括:透视投影变换、轮廓提取、特征<br/>点匹配、图像旋转、图像放缩等。实现技术要点包括;图层建立、图层移动、图层融<br/>合。论文中己给出了涉及的主要技术的原理和算法。<br/> 本课 题 作 为计算机辅助足迹检验系统的一部分，将为该系统提供更加丰富可靠的足<br/>迹图像，使该系统能更加充分的利用犯罪现场的足迹信息资源。与此同时，考虑到图像<br/>拼接技术的应用领域广泛，在课题的分析、设计等过程中，充分考虑到残缺足迹图像拼<br/>接的同时，也考虑到该技术可以应用于足迹检验以外的其他痕迹图像分析系统。<br/><br/>关键词:刑事现场，图像拼接，足迹图像<br/>2008-10-11 13:56:48 http://www.netyi.net/training/aac95266-a29e-47fc-b01f-b380ee451858论 文 题　目：中医专家系统的设计与应用<br/>学 科 名　称： 计算机应用技术 <br/><br/><br/>目 录<br/><br/>声明<br/>英文摘要<br/>摘要<br/>第一部分前言<br/><br/>1.1 HIS简介<br/>1.2中医专家系统的现状及发展趋势<br/><br/>第二部分中医专家系统及文本工作<br/><br/>2.1课题的研究意义<br/>2.2系统的功能、作用及特点<br/>2.3论文的研究内容<br/><br/>第三部分中医专家系统的开发方案<br/><br/>3.1系统开发需求分析<br/>3.2系统目标<br/>3.3系统的总体方案<br/>3.4系统组成结构<br/><br/>第四部分中医科中医专家系统的实现<br/><br/>4.1系统功能模块<br/>4.2数据库的操作<br/>4.3系统表的设计<br/>4.4系统程序设计<br/>4.5系统各个功能实现界面及部分源代码<br/><br/>第五部分面向对象技术与安全性能考虑<br/><br/>5.1面向对象技术<br/>5.2面向对象的系统分析和设计<br/>5.3面向对象实现<br/>5.4安全性能考虑<br/><br/>第六部分总结与展望<br/><br/>6.1系统实施后的效果<br/>6.2后期展望<br/><br/>附录<br/>致谢<br/>参考文献<br/><br/><br/>摘要<br/> 随着 西 南 医院信息化建设的全面发展，诊疗业务的不断拓展，以及不断递增的各<br/>类数据，建立新的中医管理体制和方法，成为中医科现代化建设的一个重要内容。为<br/>了合理利用中医科的现有资源，提升中医科的形象，我们结合西南医院中医科的实际，<br/>自行设计开发了这套中医专家系统。该系统主要由方剂与中药查询数据库、处方分析<br/>系统、辅助诊疗系统、中医文献管理和系统管理这五大功能组成。减轻了中医科工作<br/>人员的劳动强度，提高了工作效率，从而使科室能够以少的投入获得更好的社会效益<br/>和经济效益。<br/> 中医 专 家 系统是一种融合中医学、管理科学、信息科学、系统科学、电子计算机<br/>技术为一体的综合性先进管理系统。近年来，随着信息技术、计算机技术及管理技术<br/>的进步，中医专家系统在理论上和开发方式上取得了巨大发展，其应用领域也越来越<br/>广泛，这都为中医科医生工作的信息集成、信息共享提供了保障，促进了管理效率的<br/>提高。<br/> 本文 叙 述 了中医专家系统的设计思路与实现方法。主要介绍了以下几个方面:<br/>. 概述:描述了医院信息系统、中医科信息化的现状和在西南医院开发中医专家<br/>系统 的 必 要 性 。<br/>. 开发方案:介绍了开发本系统的软、硬件方案、开发工具和数据库系统。<br/>. 模块结构及实现方法:介绍了本系统各功能模块、实现这些功能的方法和过程。<br/>. 主要技术:阐述了AJAX技术和面向对象技术，并分析了系统的安全性能。<br/>. 以后的工作:总结全文并展望今后的工作方向。<br/>关键词:医院信息系统、专家系统、AJAX, C#.NET、数据库、面向对象技术<br/><br/>2008-10-11 13:52:43 http://www.netyi.net/training/ce3d8a19-8bd8-4e2b-a249-c2358e4e6ee9论 文 题　目：基于数据场理论的人脸识别系统<br/>学 科 名　称：软件工程 <br/><br/>目 录<br/><br/>摘要<br/>英文摘要<br/>第1章绪论<br/>第2章人脸识别理论与技术<br/>第3章人脸检测与图像归一化<br/>第4章人脸图像的特征提取<br/>第5章分类器设计<br/>第6章软件系统的组成和操作<br/>第7章全文总结与展望<br/>致谢<br/>参考文献<br/><br/>摘要<br/>人脸 识 别 因其在安全验证系统、信用卡验证、医孚、档案管理、视频会议、<br/>人机交互、系统公安(罪犯识别等)等方面的巨大应用前景，越来越成为当前模式<br/>识别和人工智能领域的一个研究热点。然而，尽管人类能毫不费力地识别出人脸<br/>及其表情，人脸的自动机器识别却是一个难度极大的课题。面对一张数字化的二<br/>维矩阵，机器本身不具备识别的能力。人脸识别技术牵涉到模式识别、图像处理<br/>及生理、心理学等方面的诸多知识。随着社会生活等各方面数字化的推进，人脸<br/>识别在安全方面的应用已经越来越受到人们的注意，特别是在近年来推出的数字<br/>家庭概念中，人脸识别在家庭安全中的应用具有独到的优势。<br/>本文 的 研 究内容主要是单人正面人脸图像的特征提取以及鉴别和匹配。论文<br/>描述了人脸识别的主要过程:人脸检测和图像初始化、人脸特征的提取、特征的<br/>鉴别和匹配。首先利用灰度在垂直和水平方向的积分投影曲线，确定人眼的位置，<br/>然后将图像以瞳孔之间的间距为基准，按一定的比例进行归一化，获得了人脸表<br/>示的几何不变性。在特征提取阶段，利用了基于数据场的特征提取算法，该方法<br/>将人脸图像的像素点作为数据点，像素值作为数据点的值，形成数据场，提取数<br/>据场中的势值较大的点的位置和大小作为特征向量，用于匹配识别。然后，将数<br/>据场思想应用到分类鉴别的过程中。聚类算法通过数据形成数据场中等势线的嵌<br/>套，自然地完成数据的聚类。而数据场中势值最低的点，就是最离群的点。该算<br/>法以可视化的方法处理数据，过程直观形象，比常规的方法更有表现力。在<br/>Microsoft windowsXP平台上VC + 6.0开发环境下，采用上述算法编制了人脸<br/>识别的软件。<br/>2008-10-11 13:48:24 http://www.netyi.net/training/b9541abb-22d4-477d-80d2-be9f82c43d58论 文 题　目：基于GIS的道路交通安全评价研究<br/>学 科 名　称：交通运输规划与管理 <br/><br/>目 录<br/><br/>文摘<br/>英文文摘<br/>声明<br/>第一章绪论<br/><br/>1.1课题来源<br/>1.2问题的提出<br/>1.3国内外相关研究概况<br/>1.3.1道路交通安全评价方面<br/>1.3.2 GIS技术在交通安全中的应用<br/>1.4研究目标与主要内容<br/>1.4.1研究目标<br/>1.4.2主要研究内容<br/>1.5研究技术路线<br/>第二章道路交通安全评价体系研究<br/>2.1概述<br/>2.1.1道路交通安全评价目的与意义<br/>2.1.2道路交通安全评价层次划分<br/>2.1.3建立评价体系的步骤<br/>2.2影响道路交通安全的因素<br/>2.3道路交通安全评价指标建立<br/>2.3.1评价指标的选取原则<br/>2.3.2道路交通安全评价指标<br/>2.4道路交通安全评价方法构建<br/>2.4.1已有交通安全评价方法综述<br/>2.4.2区域交通安全评价<br/>2.4.3路网交通安全评价<br/>2.4.4路段或地点交通安全评价<br/>2.5本章小结<br/>第三章RTSE-GIS系统体系结构分析与设计<br/>3.1地理信息系统与交通安全分析<br/>3.1.1地理信息系统简介<br/>3.1.2 GIS-T在交通安全分析中的功能<br/>3.2系统需求分析<br/>3.2.1用户群及其需求<br/>3.2.2功能需求<br/>3.2.3数据需求<br/>3.2.4性能需求<br/>3.2.5运行需求<br/>3.3系统体系结构设计<br/>3.3.1 GIS设计理论<br/>3.3.2系统总体结构设计<br/>3.4子系统与功能模块划分<br/>3.4.1数据的管理维护子系统<br/>3.4.2事故查询统计子系统<br/>3.4.3安全分析与评价子系统<br/>3.4.4电子地图操作<br/>3.4.5系统设置<br/>3.5系统开发模式选择<br/>3.5.1应用型GIS开发的三种实现模式<br/>3.5.2开发模式的比较确定<br/>3.6本章小结<br/>第四章RTSE-GIS系统数据库技术<br/>4.1系统数据库需求分析<br/>4.1.1系统数据库的地位及设计原则<br/>4.1.2系统数据库的数据分析<br/>4.2基于多维数据管理技术的交通安全数据库<br/>4.2.1多维数据模型<br/>4.2.2多维数据模型的交通事故数据库<br/>4.3 GIS与交通安全数据整合技术<br/>4.2.1 GIS-T线性数据处理相关技术<br/>4.2.2数据整合规划<br/>4.2.3建立LRS事故数据与道路静态数据<br/>4.2.4建立GIS路网数据<br/>4.2.5数据存储和访问<br/>4.4本章小结<br/>第五章RTSE-GIS系统功能设计与开发<br/>5.1系统开发流程及环境<br/>5.1.1开发流程<br/>5.1.2开发环境<br/>5.2电子地图操作<br/>5.3数据管理和维护<br/>5.3.1事故点信息录入和修改<br/>5.3.2空间数据与属性数据绑定<br/>5.4事故信息查询统计<br/>5.5基于GIS的交通安全分析与评价<br/>5.5.1区域交通安全评价<br/>5.5.2路网交通安全评价分析<br/>5.5.3地点或路段事故黑点分析<br/>5.5.4功能实现<br/>5.6本章小结<br/>第六章结论与展望<br/>6.1主要研究成果<br/>6.2主要创新点<br/>6.3有待进一步研究的问题<br/>致谢<br/>参考文献<br/><br/><br/>摘要<br/> 本学 位 论 文是结合国家自然科学基金联合资助项目《公路交通事故黑点分析技术研究》和江苏<br/>省交通厅科技项目《一级公路设计安全性评价方法研究》完成的。<br/> 随着 我 国 道路通车里程进一步增加，机动车保有R的快速发展，道路交通安全的改善显得越发<br/>重要。如何有效改善道路交通安全现状，建立一种适合道路交通安全分析和评价的道路交通安全信<br/>息系统势在必行。因此.本文结合我国道路交通实际。建立一套道路交通安全评价体系。并构建基<br/>于GIs技术的道路交通安全评价系统，为道路现状的交通安全分析和评价提供帮助和指导。<br/>首先 ， 在 对我国道路交通安全影响因素分析和安全评价需求分析的基础七，按被比较对象的空<br/>间尺度大小，将道路交通安全评价划分为面、线、点三个层次进行评价，即试域交通安全评价、路<br/>网交通安全评价、路段或地点交通安全评价，筛选出评价指标.确定相应评价模型和方法，构建道<br/>路交通安全评价体系。<br/> 其次 ， 进 行系统用户需求、功能需求和数据需求等需求分析，根据系统的需求分析，提出了基<br/>于GIs的道路交通安全评价系统体系结构，整个系统由三个层次组成:数据管理层、中间技术层和<br/>应用服务层。进而确定了系统的主要功能，包括:电子地图操作和系统设置等基本功能和数据的管<br/>理维护、事故查询统计分析、交通安全分析与评价等安全分析业务功能。最后通过对比GIs应用开<br/>发模式，确定选用集成二次开发方式，并以GIs组件技术对系统进行开发。<br/> 第三 ， 分 析道路交通安全信息构成，利用数据挖掘的多维数据管理技术，以关系V?数据库MSS QL<br/>Server为平台，建立了适合交通安全分析与评价研究的交通事故的多维数据概念模型，由交通事故<br/>信息事实表和事故特征维表、事故位置维表、当事人维表、事故车辆维表、事故时间维表和交通环<br/>境维表6个维表构成。提出了GIs与交通安全数据的整合技术，对GIS-T线性数据处理相关技术线<br/>性参考方法、动态分段，GIs路网数据库的建立及数据存储和访问分别进行详细探讨，实现了系统<br/>空间数据和外部信息数据库的动态链接。<br/> 最后 . 确 定了系统的研发技术路线与运行开发环境，对系统各功能模块进行了详细研究和阐述。<br/>在道路交通安全评价分析方法的基础上，结合GIs的信息显示与查询、空间实体的合并与分割、缓<br/>冲区分析、动态分段、线性参考、专题图等功能，提出了基于GIs的区域、路网、路段或地点的交<br/>通安全分析和评价方法。利用GIs技术和系统集成技术，及相关数理统计分析算法，研发实现了基<br/>于GIs的道路交通安全评价的实验性系统。<br/><br/>关键 词 : 地理信息系统(GIs)、道路交通安全、安全评价、多维信息管理、数据库技术、线性<br/>参考、动态分段<br/><br/><br/>2008-10-11 13:41:35 http://www.netyi.net/training/53c62b3a-adfd-434e-bd8d-71397360dc93论 文 题　目：基于GIS的物流配送系统的研究与设计<br/><br/>学 科 名　称：计算机应用技术 <br/><br/>摘 要<br/><br/> 现代物流作为一种先进的组织方式和管理技术，通过降低流通费用，缩短流<br/>通时间，可以整合企业价值链、延伸企业的控制能力，加快企业资金周转，从而<br/>成为企业“第三利润源”。特别是在电子商务环境下，供应商必须全面及时地掌<br/>握物流各个环节即时信息，以此制定生产和销售计划，及时调整市场策略。把地<br/>理信息系统（GIS）技术融入到物流配送的过程中，就可以更容易地处理物流配<br/>送中的各个环节，并对其中涉及地理信息的，诸如物流设施定位、运输车辆的调<br/>度和配送路线的选择、最优库存控制等问题进行有效管理和决策分析，有助于物<br/>流配送企业有效地利用现有资源，降低消耗，提高效率。<br/> 在物流配送管理中，合理选择配送路线是有效控制物流成本的关键。本文主<br/>要针对城市物流配送的特点，结合实际应用，建立车辆路径优化模型，并对基于<br/>GIS的物流配送系统集成的关键技术进行了研究。文中首先介绍了国内外基于GIS<br/>物流配送系统的研究现状，接着对配送中所需考虑的核心问题---配送车辆路径<br/>规划进行了研究，运用贪婪算法将VRP问题进行分步求解，使之成为求K条路线的<br/>TSP，再引入改进蚁群算法来解决TSP问题。并通过仿真实验，研究分析了蚁群算<br/>法在求解中小规模的TSP问题上，各参数的设置对算法的性能的影响。在上述实<br/>验结果的基础上，将该算法求解模型引入物流配送决策系统，初步搭建物流配送<br/>决策系统，模拟城市流配送优化决策。最后，通过系统流程的分析，提出切合实<br/>际系统集成方案，对数据集成与功能集成进行了深入探讨，实现了空间和属性数<br/>据库的统一存储管理、电子地图基本操作功能、最短路径查找功能组件模块。文<br/>章最后通过一个模拟算例，将本文提出的优化模型与已有优化算法进行了比较，<br/>进一步验证了本文给出的模型的优越性。<br/><br/>关键字： GIS，物流配送，路径优化，蚁群算法，TSP<br/><br/>目 录<br/><br/>文摘<br/>英文文摘<br/>第一章 绪论<br/>第二章 物流配送系统规划<br/>第三章 物流配送车辆路径规划数学模型及求解算法<br/>第四章 地理信息系统（GIS）引入物流配送<br/>第五章 基于GIS物流配送系统集成开发<br/>第六章 总结与展望<br/>致谢<br/>参考文献2008-10-11 13:31:12 http://www.netyi.net/training/089ffec8-cecc-4488-8750-b487930b7800Many professional people in science and technology have an excellent command of their scientific subject, but have difficulty in expressing their knowledge in simple, accurate English. Good Style explains the tactics that can be used to express scientific information in a coherent, readable style. The book discusses in detail possible choices of vocabulary,phrasing and sentence structure. The advice offered is based on evidence of the styles preferred by professional readers. Each piece of advice is supported by many examples of writing from a variety of scientific and technical contexts.<br/>John Kirkman draws from his many years of experience lecturing on communication<br/>studies in Europe, the USA, the Middle East, and Hong Kong, both in academic<br/>programmes and in courses for large companies, research centres and government<br/>departments.<br/>Good Style has become a standard reference book on the shelf of students of science,technology, medicine, and computing, and is an essential aid to all professionals whose work involves writing of reports, papers, guides, manuals, or on-screen texts (including material to be presented on the web). This new edition includes additional examples of how to express medical and life-science information.<br/>John Kirkman now works as a consultant specialising in research and training in<br/>scientific and technical communication. Previously he was Director of the<br/>Communication Studies Unit at the University of Wales, Cardiff.2008-09-27 23:29:05 http://www.netyi.net/training/87bd6d6c-5321-4c53-8980-cd1571c83b65清华大学硕士学位论文<br/><br/>多传感器数据融合发展到今天，其研究横跨信息科学的多个学科，它的应<br/>用覆盖众多应用领域。如何利用数据融合来改善多传感器系统的性能，已经成<br/>为目前学术界和工业界面临的一个重要问题。无论是抽象与概括多传感器数据<br/>融合通用的模式与方法，还是面向应用设计与实现具体多传感器数据融合系统，<br/>都从属于多传感器数据融合研究的范畴。这些研究可归结为系统设计和算法设<br/>计两个方面。<br/>本文在系统设计方面回顾并总结了多传感器数据融合的一般模式、分层模<br/>型、融合结构等，在算法设计方面介绍了确定性目标动态模型下与非确定性目<br/>标动态模型下两类航迹融合算法。通过比较说明了分布式系统结构和异步航迹<br/>融合在多传感器多目标跟踪系统中的重要性和现实意义。<br/>从系统特征入手，我们构建了一种高性能航迹融合反馈系统模型。本文对<br/>系统模型的设计策略、系统结构和相应算法给予了详细描述和说明。该模型在<br/>功能与结构上实现了误差解耦，功能模块解耦，异步数据处理，以及系统闭环<br/>控制，比传统航迹融合系统具有更好的稳定性、可靠性和可扩展性。多传感器<br/>多目标场景下的仿真测试，显示了该系统在误差校正、航迹关联、航迹融合方<br/>面的良好性能。<br/>上述航迹融合反馈系统采用了基于协方差交集（Covariance Intersection, CI）<br/>的异步航迹融合算法，简记为CI算法。为提高高速机动目标环境中的航迹融合<br/>性能，本文在该算法基础上提出了一种改进的航迹融合算法：航迹片段协方差<br/>交集异步航迹融合算法，简记为Segment-CI 算法。该算法采用最小二乘曲线拟<br/>合技术将目标航迹实时划分为直线段和圆弧段，在不同航迹片段上采用不同的<br/>预测模型，使用历史数据提高状态预测精度，沿用CI技术合并局部状态估计。<br/>异步多传感器环境下的仿真测试中，Segment-CI与CI两种算法的位置、速度估<br/>计误差均方根比值均小于1，表明该算法的航迹融合性能优于CI算法。<br/>本文的研究工作是在多传感器数据融合系统设计与算法改进方面有益的尝<br/>试，其结论对于类似系统与算法具有一定的借鉴意义。2008-09-17 19:07:44 http://www.netyi.net/training/981c8f1e-95cc-4227-b89d-930a1760676c论 文 题　目：基于纹理分析的图像检索<br/>学 科 名　称：通信与信息系统<br/>摘　　　　要：<br/> 纹理检索是基于内容的图像检索的重要组成部分，而纹理特征的提取与分析性能直接关系到纹理检索的性能，是纹理检索的研究重点。本文主要研究用于图像检索的纹理分析算法，主要工作如下：<br/> (1)本文在分析总结现有纹理分析算法的基础上构造窗口纹理分析方法，用于纹理特征的提取与分析。该方法综合了统计与结构两类纹理分析方法的思想，利用结构法的纹理基元分析思想和统计法的整体特征统计构造新的纹理特征提取思想，并利用该思想分别在空间域、频率域及多分辨率基础上进行了纹理分析算法的设计与实验；<br/> (2)在空间域上，利用窗口法思想设计出效率较高的差分矩阵方法和能对纹理基元进行精确分析的质心模式统计方法；<br/> (3)在频率域上，通过对频域窗口的分析与改进，设计出小波频谱系数的纹理分析方法，该方法同时具有小波的多尺度特性和傅立叶变换的自配准性质；<br/> (4)在纹理的多分辨率分析的基础上，本文将窗口法与时/频域分析的小波包变换进行了结合，设计出多分辨率的差分矩阵与质心模式统计，改进了差分矩阵与质心模式统计的检索性能，证明了窗口法与时/频域分析结合的可行性。<br/> (5)利用Brodatz自然纹理库形成的两个测试集对本文所设计的五种纹理分析算法进行了综合的比较和实验，总结出各自的优缺点及适用场合并探讨了窗口纹理分析方法的研究方向，为日后的研究奠定一定的基础。<br/>目 录<br/><br/>文摘<br/>英文文摘<br/>第1章绪论<br/>第2章纹理分析算法与相关技术<br/>第3章差分矩阵与质心模式统计<br/>第4章小波系数频谱算法<br/>第5章多分辨率差分矩阵与质心模式统计<br/>第6章算法综合性能分析与比较<br/>结论<br/>参考文献<br/><br/>2008-09-05 11:14:16 http://www.netyi.net/training/b454004f-d2bd-4e6e-9439-19ccc49de1a5论 文 题　目：基于小波变换的图像检索技术研究<br/>摘　　　　要：<br/> 图像检索是多媒体应用的关键技术，特别是基于压缩域图像检索技术，由于处理的数据量少，减少了对计算机资源的需求，提高了系统的实时性、高效性和灵活性，对多媒体业务的普及和拓展以及多媒体网络的繁荣起着积极地推动作用。因此基于内容的压缩域图像检索有着重要的理论意义和应用价值，是当前图像检索的热点和趋势。本文对基于压缩域的图像检索进行了研究，主要工作如下：<br/> (1)系统地论述了压缩域图像检索技术的发展现状，总结了图像检索的一些关键技术。<br/> (2)在对子带小波系数分布的统计特性进行分析的基础上，研究了基于广义高斯分布的小波域图像检索方法，实验发现：对于系数动态范围大而峰值又尖锐的图像，广义高斯分布不能够很好地描述这种小波系数的分布，这会影响到图像检索的效果。由于高斯混合模型具有一个重要特性：如果模型中的成员足够多，就可以任意精度地逼近任意的连续分布。因此本文提出改用高斯混合模型来拟合子带小波系数的分布。实验显示，混合模型对小波系数分布的拟合效果比广义高斯分布有明显提高，证明改进的想法是可行的。<br/> (3)对基于广义高斯分布的小波域图像检索方法进行改进。采用两个分量的高斯混合模型拟合小波系数的统计直方图，近似描述系数的分布情况。通过EM算法估算出模型的参数，根据高斯混合模型的相似性匹配进行图像的检索。实验证明，采用高斯混合模型的方法比采用广义高斯分布的方法，检索准确率有一定的提高。<br/>目 录<br/><br/>文摘<br/>英文文摘<br/>西北大学学位论文知识产权声明书及西北大学学位论文独创性声明<br/>第一章绪论<br/>第二章图像检索的关键技术<br/>第三章基于广义高斯分布的小波域图像检索方法<br/>第四章基于高斯混合模型的小波域图像检索方法研究<br/>第五章总结与展望<br/>参考文献<br/>致谢2008-09-05 11:08:51 http://www.netyi.net/training/fea292a0-273d-436d-8994-963f9e70c008大连理工大学<br/>硕士学位论文--基于内容的图像检索技术研究<br/><br/>　目 录<br/><br/>文摘<br/>英文文摘<br/>独创性说明及大连理工大学学位论文版权使用授权书<br/>引 言<br/>1基于颜色特征的图像检索<br/>1.1基于颜色的特征表达<br/>1.1.1颜色模型<br/>1.1.2颜色特征的表达<br/><br/>1.2颜色特征的匹配方法<br/><br/>1.2.1五种简单的直方图匹配方法<br/>1.2.2基于局部累加直方图的两种检索算法<br/><br/>2基于纹理特征的图像检索<br/><br/>2.1概述<br/>2.2纹理特征的匹配方法<br/><br/>2.2.1统计法纹理描述<br/>2.2.2频谱法纹理描述<br/><br/>3基于形状特征的图像检索<br/><br/>3.1概述<br/>3.2形状特征的匹配方法<br/><br/>3.2.1基于特征的方法<br/>3.2.2基于变换域的方法<br/>3.2.3基于变形的方法<br/>3.2.4草图查询<br/><br/>4本文的检索算法<br/><br/>4.1颜色特征提取模块<br/>4.2结构特征提取模块<br/>4.3检索效果<br/><br/>5本文改进的算法及新的检索技术<br/><br/>5.1高效的图像检索相似度量<br/><br/>5.1.1 非相似距离<br/>5.1.2相似距离<br/>5.1.3 E+吉布斯分布<br/>5.1.4图像检索效果<br/>5.1.5实证分析<br/><br/>5.2新的图像检索技术<br/><br/>5.2.1扩展的数据库技术<br/>5.2.2基于PSO的最大类间方差图像分割方法<br/><br/>结 论<br/>参考文献<br/>攻读硕士学位期间发表学术论文情况<br/>致 谢<br/>2008-09-02 16:28:26 http://www.netyi.net/training/5b857508-196d-44b9-b82c-8e904c466b80个文共分5章。第」章足绪沦，介绍了选题的背景，捉{日了沦文的主要研究<br/>内容第一章介绍了内千可航道几维:tJ’视化的关键技术，包括人比例尺电子航道图<br/>数据的提取和预处理方法，航道地形、助航标志以及航道两侧建筑物的三维建模<br/>力一法〕第二章介绍了内河航道空间数据库的设计与实现方一法，以节从空间数据库<br/>的丰既念入乒，探讨了利用FSRIG。、，database空间数据模型设计和实现内河航道空<br/>阳{数扣{)车l狗方法。第四章给出了内河航道三维显不与分析系统I狗)1发方法，首先<br/>介绍了Al’cGISFngille开发平台，然后给出了系统的总体结构设计，最后介绍了<br/>各个功能模块的实现方法。第山帝给出了论文的结论并对进一步的}一作进行了展<br/>望2008-08-20 07:57:49 http://www.netyi.net/training/d696c454-b688-47d8-8b72-dbc24fc6767f一　　导论<br/>二　　纯认识的兴趣之存在与理及概念<br/>三　　普遍之理与无私之心<br/>四　　经验知识中之法执及其解脱之历程<br/>五　　推理之知识中之互证与知识之形式之法执<br/>六　　知识之经验内容之超越与逻辑学思维<br/>七　　形数之普遍性与法执之解脱<br/>八　　理解历史事物之心之超越性<br/>九　　应用科学知识之心灵之涵盖性<br/>十　　科学意识之道德价值及逻辑意识哲学意识<br/>十一　与一般科学意识关联之哲学意识---科学的宇宙观意识及知识论意识<br/>十二　形上学意识<br/>十三　道德哲学之意识<br/>十四　文化哲学与历史哲学意识<br/>十五　求真理心之道德性与其退堕<br/>十六　自陷于已成知识之心态与虚心求真理之心态之涵义2008-08-18 23:56:54 http://www.netyi.net/training/a77ee9dd-c2c2-41f0-88f0-9f33f345d380In this article we will compare the classical<br/>methods of Fourier analysis with the newer<br/>methods of wavelet analysis. Given a signal,<br/>say a sound or an image, Fourier analysis<br/>easily calculates the frequencies and<br/>the amplitudes of those frequencies which make<br/>up the signal. This provides a broad overview of<br/>the characteristics of the signal, which is important<br/>for theoretical considerations. However,<br/>although Fourier inversion is possible under<br/>certain circumstances, Fourier methods are not<br/>always a good tool to recapture the signal, particularly<br/>if it is highly nonsmooth: too much<br/>Fourier information is needed to reconstruct<br/>the signal locally. In these cases, wavelet analysis<br/>is often very effective because it provides a<br/>simple approach for dealing with local aspects<br/>of a signal. Wavelet analysis also provides us with<br/>new methods for removing noise from signals<br/>that complement the classical methods of<br/>Fourier analysis. These two methodologies are<br/>major elements in a powerful set of tools for theoretical<br/>and applied analysis.<br/>This article contains many graphs of discrete<br/>signals. These graphs were created by the computer<br/>program FAWAV, A Fourier–Wavelet Analyzer,<br/>being developed by the author.2008-08-09 11:17:17 http://www.netyi.net/training/9740383f-00bc-4def-823c-3a679ed5e839The Econometrics of Option Pricing&#164;<br/>Ren?e Garcia<br/>Universit?e de Montr?eal, CIRANO, CIREQ<br/>Eric Ghysels<br/>University of North Carolina and CIRANO<br/>?E<br/>ric Renault<br/>Universit?e de Montr?eal, CIRANO, CIREQ<br/>First draft: November 2001<br/>This version: August 1, 2003<br/>Keywords: Stock PriceDynamics, Multivariate Jump-Di?usionModels, Latent variables, Stochastic<br/>Volatility, Objective and Risk Neutral Distributions, Nonparametric Option Pricing, Discretetime<br/>Option Pricing Models, Risk Neutral Valuation, Preference-free Option Pricing.<br/>JEL Classiˉcation: C1,C5,G1<br/>&#164;Address for correspondence: D?epartement de Sciences ?E conomiques, Universit?e de Montr?eal, C.P.<br/>6128, Succ. Centre-Ville, Montr?eal, Qu?ebec, H3C 3J7, Canada. The ˉrst and the last authors gratefully<br/>acknowledge ˉnancial support from the Fonds de la Formation de Chercheurs et l'Aide μa la Recherche du<br/>Qu?ebec (FCAR), the Social Sciences and Humanities Research Council of Canada (SSHRC), the Network<br/>of Centres of Excellence MITACS and the Institut de Finance Math?ematique de Montr?eal (IFM2). The<br/>second author thanks CIRANO for ˉnancial support.<br/>1 Introduction and overview<br/>The growth of the option pricing literature parallels the spectacular developments of derivative<br/>securities and the rapid expansion of markets for derivatives in the last three decades.<br/>Writing a survey of option pricing models appears therefore like a formidable task. To<br/>delimit our focus we will put emphasis on the more recent contributions since there are<br/>already a number of surveys that cover the earlier literature. For example, Bates (1996b)<br/>wrote an excellent review, discussing many issues involved in testing option pricing models.<br/>Ghysels, Harvey and Renault (1996) and Shephard (1996) provide a detailed analysis of<br/>stochastic volatility modelling, while Renault (1997) explores the econometric modelling<br/>of option pricing errors. More recently, Sundaresan (2000) surveys the performance of<br/>continuous-time methods for option valuation. The material we cover obviously has many<br/>seminal contributions that pre-date the most recent work. Needless to say that due credit<br/>will be given to the seminal contributions related to the general topic of estimating and<br/>testing option pricing models. A last introductory word of caution: our survey deals almost<br/>exclusively with studies that have considered modelling the return process of a stock index<br/>and determining the price of European options written on this index.<br/>One of the main advances that marked the econometrics of option pricing in the last<br/>ˉve years has been the use of price data on both the underlying asset and options to jointly<br/>estimate the parameters of the process for the underlying and the risk premia associated<br/>with the various sources of risk. Even if important progress has been made regarding<br/>econometric procedures, the lesson that can be drawn from the numerous investigations,<br/>both parametric and nonparametric, in continuous time or in discrete time, is that the<br/>empirical performance still leaves much room for improvement. The empirical option pricing<br/>literature has revealed a considerable divergence between the risk-neutral distributions<br/>estimated from option prices after the 1987 crash and conditional distributions estimated<br/>from time series of returns on the underlying index. Three facts clearly stand out. First,<br/>the implied volatility extracted from at-the-money options di?ers substantially from the<br/>realized volatility over the lifetime of the option. Second, risk neutral distributions feature<br/>substantial negative skewness which is revealed by the asymmetric implied volatility curves<br/>when plotted against moneyness. Third, the shape of these volatility curves changes over<br/>time and maturities, in other words the skewness and the convexity are time-varying and<br/>maturity-dependent. Our survey will therefore explore possible explanations for the divergence<br/>between the objective and the risk neutral distributions. Modelling of the dynamics<br/>of the underlying asset price is an important part of the puzzle, while another essential<br/>element is the existence of time-varying risk premia. The last issue stresses the potentially<br/>1<br/>explicit role to be played by preferences in the pricing of options, a departure from the<br/>central tenet of the preference-free paradigm.<br/>The main approach to modelling stock returns at the time prior surveys were written,<br/>was a continuous time stochastic volatility (henceforth SV) di?usion process possibly<br/>augmented with an independent jump process in returns. Heston (1993) proposed a SV<br/>di?usion model for which one could derive analytically an option pricing formula. Soon<br/>thereafter, see e.g. Du&#177;e and Kan (1996), it was realized that Heston's model belonged to<br/>a rich class of a&#177;ne jump di?usion processes for which one could obtain similar results.<br/>Du&#177;e, Pan and Singleton (2000) discuss equity and ˉxed income derivatives pricing for<br/>the general class of a&#177;ne jump di?usions. The evidence regarding the empirical ˉt of the<br/>a&#177;ne class of processes is mixed, see e.g. Dai and Singleton (2000), Chernov, Gallant,<br/>Ghysels and Tauchen (2003), Ghysels and Ng (1998) for further discussion. There is a<br/>consensus that single volatility factor models, a&#177;ne (like Heston (1993)) or non-a&#177;ne (like<br/>Hull and White (1987) or Wiggins (1987)), do not ˉt the data (see Andersen, Benzoni and<br/>Lund (2002), Benzoni (1998), Chernov, Gallant, Ghysels and Tauchen (2003), Pan (2002),<br/>among others). How to expand single volatility factor di?usions to mimic the data generating<br/>process remains unsettled. Several authors augmented a&#177;ne SV di?usions with jumps,<br/>see Andersen, Benzoni and Lund (2001), Bates (1996a), Chernov, Gallant, Ghysels and<br/>Tauchen (2003), Eraker, Johannes and Polson (2001), Pan (2002), among others. Bakshi,<br/>Cao and Chen (1997), Bates (2000) Chernov, Gallant, Ghysels and Tauchen (2003) and<br/>Pan (2002) show, however, that SV models with jumps in returns are not able to capture all<br/>the empirical features of observed option prices and returns. Bates (2000) and Pan (2002)<br/>argue that the speciˉcation of the volatility process should include jumps, possibly correlated<br/>with the jumps in returns. Chernov, Gallant, Ghysels and Tauchen (2003) maintain<br/>that a two-factor non-a&#177;ne logarithmic SV di?usion model without jumps yields a superior<br/>empirical ˉt compared to a&#177;ne one-factor or two factor SV processes, or SV di?usions<br/>with jumps. Alternative models were also proposed in recent years: they include volatility<br/>models of the Ornstein-Uhlenbeck type but with L?evy innovations (Barndor?-Nielsen and<br/>Shephard, 2001) and stochastic volatility models with long memory in volatility (Breidt,<br/>Crato and de Lima (1998)) and Comte and Renault (1998)).<br/>The statistical ˉt of the underlying process and the econometric complexities associated<br/>with it should not be the only concern, however. An important issue for option pricing is<br/>whether or not the models deliver closed-form solutions. We will therefore discuss if and<br/>when there exists a trade-o? between obtaining a good empirical ˉt or a closed-form option<br/>pricing formula.The dynamics of the underlying fundamental asset cannot be related to<br/>option prices without additional assumptions or information. One possibility is to assume<br/>2<br/>that the risks associated with stochastic volatility or jumps are idiosyncratic and not priced<br/>by the market. There is a long tradition of this, but most recent empirical work clearly<br/>indicates there are prices for volatility and jump risk (see e.g. Andersen, Benzoni and Lund<br/>(2002), Chernov and Ghysels (2000), Pan (2002), among others). One can simply set values<br/>for these premia and use the objective parameters to derive implications for option prices as<br/>in Andersen, Benzoni and Lund (2001). A more informative exercise is to use option prices<br/>to calibrate the parameters under the risk neutral process given some version of a nonlinear<br/>least-squares procedure as in Bakshi, Cao and Chen (1997) and Bates (2000). An even more<br/>ambitious program is to use both the time series data on stock returns and the panel data<br/>on option prices to characterize the dynamics of returns with stochastic volatility and with<br/>or without jumps as in Chernov and Ghysels (2000), Pan (2002), Poteshman (2000) and<br/>Garcia, Lewis and Renault (2001).<br/>Whether one estimates the objective probability distribution, the risk neutral or both,<br/>there are many challenges in estimating the parameters of di?usions. The presence of latent<br/>volatility factors make maximum likelihood estimation computationally infeasible. This is<br/>the area where probably the most progress has been made in the last few years. Several<br/>methods have been designed for the estimation of continuous time dynamic state-variable<br/>models with the pricing of options as a major application. Simulation-based methods have<br/>been most successful in terms of empirical implementations. That includes the indirect<br/>inference and e&#177;cient methods of moments of Gouri?eroux, Monfort and Renault (1993)<br/>and Gallant and Tauchen (1996) respectively, and several procedures discussed by Johannes<br/>and Polson (2002) as well as A?3t-Sahalia, Hansen and Scheinkeman (2002) in thisHandbook.<br/>Another approach is to use implied state methods. While Pastorello, Patilea and Renault<br/>(2003) base an indirect inference approach on Black-Scholes implied volatilities, Pan (2002)<br/>uses the Fourier transform to derive a set of moment conditions pertaining to implied states.<br/>Renault and Touzi (1996), Patilea and Renault (1997) and Renault (1997) propose iterative<br/>and recursive procedures which extend the EM (expectation-maximization) methodology<br/>to maximum likelihood contexts where it usually does not apply. Pastorello, Patilea and<br/>Renault (2003) propose a general methodology of iterative and recursive estimation in<br/>structural non-adaptive models which nests all the previous implied state approaches.<br/>Nonparametric methods have also been used extensively. Several studies aimed at<br/>recovering the risk-neutral probabilities or state-price densities implicit in option or stock<br/>prices. For instance, Rubinstein (1996) proposed an implied binomial tree methodology to<br/>recover risk-neutral probabilities which are consistent with a cross-section of option prices.<br/>A?3t-Sahalia and Lo (1998) use a kernel estimator of the volatility function in a Black-<br/>Scholes type model. Stutzer (1996) uses an approach called canonical valuation which<br/>3<br/>uses past return data and possibly but not necessarily option price data to estimate the<br/>payo? distribution at expiration. Another approach consists in estimating directly the<br/>option pricing function with nonparametric methods. Hutchinson, Lo and Poggio (1994),<br/>Broadie, Detemple, Ghysels and Torrμes (2000a,b), and Garcia and Gen?cay (2000) follow<br/>this route. An important issue with the model-free nonparametric approaches is that the<br/>recovered risk-neutral probabilities are not always positive and one may consider adding<br/>constraints on the pricing function or the state-price densities. For example, A?3t-Sahalia<br/>and Duarte (2003) impose monotonicity and convexity restrictions using a nonparametric<br/>method based on locally polynomial estimators.<br/>Bates (2000), among others, shows that risk-neutral distributions recovered from option<br/>prices before and after the crash of 1987 are fundamentally di?erent whereas the objective<br/>distributions do not show such structural changes. Before the crash, both the risk neutral<br/>and the actual distributions look roughly lognormal. After the crash, the risk-neutral<br/>distribution is left skewed and leptokurtic. A possible explanation for the di?erence is<br/>a large change in the risk aversion of the average investor. Since risk aversion can be<br/>recovered empirically from the risk neutral and the actual distributions, A?3t-Sahalia and<br/>Lo (2000), Jackwerth (2000) and Rosenberg and Engle (2002) estimate preferences for<br/>the representative investor using simultaneously S&amp;amp;P500 returns and options prices for<br/>contracts on the index. Preferences are recovered based on distance criteria between the<br/>model risk neutral distribution and the risk neutral distribution implied by option price<br/>data.<br/>Another approach of recovering preferences is to set up a representative agent model and<br/>estimate the preference parameters from the ˉrst-order conditions using a GMM approach.<br/>While this has been extensively done with stock and Treasury bill return data (see Hansen<br/>and Singleton (1982), Epstein and Zin (1991) among others), it is only recently that Garcia,<br/>Luger and Renault (2003) estimated preference parameters in a recursive utility framework<br/>using option prices. In this survey we will discuss under which statistical framework option<br/>pricing formulas are preference-free and risk-neutral valuation relationships (Brennan, 1979)<br/>hold in a general stochastic discount factor framework (Hansen and Richard (1987)). When<br/>these statistical restrictions do not hold, it will be shown that preferences play a role. Bates<br/>(2001) argues that the overall industrial organization of the stock index option markets is<br/>not compatible with the idealized construct of a representative agent. He therefore proposes<br/>an equilibrium analysis with investor heterogeneity.<br/>Apart from statistical model ˉtting, there are a host of other issues pertaining to the<br/>implementation of models in practice. A recent survey by Bates (2003) provides an overview<br/>of the issues involved in empirical option pricing, especially the questions surrounding data<br/>4<br/>selection, estimation or calibration of the model and presentation of results.2008-07-22 21:22:19 http://www.netyi.net/training/c40baa79-a00f-44d1-a1b9-b59aff909e6eThe Analysis of the Cross Section<br/>of Security Returns<br/><br/>Ravi Jagannathany<br/>Georgios Skoulakisz<br/>Zhenyu Wangx<br/>&amp;#x3;This paper will appear as a chapter in the forthcoming Handbook of Financial Econometrics edited by Yacine<br/>A&amp;#x10;t-Sahalia and Lars P. Hansen. The authors wish to thank Bob Korajczyk, Ernst Schaumburg and Jay Shanken for<br/>comments and Aiyesha Dey for editorial assistance.<br/>yDepartment of Finance, Kellogg School of Management, Northwestern University, Evanston, IL 60201, USA.<br/>E-mail: rjaganna@kellogg.northwestern.edu<br/>zDepartment of Finance, Kellogg School of Management,Northwestern University, Evanston, IL 60201, USA.<br/>E-mail: g-skoulakis@kellogg.northwestern.edu<br/>xDepartment of Finance and Economics, Graduate School of Business, Columbia University, New York, NY 10027,<br/>USA. E-mail: zwang@columbia.edu<br/><br/><br/>Contents<br/>1 Introduction 1<br/>2 Linear Beta Pricing Models, Factors and Characteristics 3<br/>2.1 Linear beta pricing models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4<br/>2.2 Factor selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4<br/>2.3 Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7<br/>3 Cross-Sectional Regression Methods 8<br/>3.1 Description of the CSR method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br/>3.2 Consistency and asymptotic normality of the CSR estimator . . . . . . . . . . . . . . 10<br/>3.3 Fama-MacBeth variance estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13<br/>3.4 Conditionally homoscedastic residuals given the factors . . . . . . . . . . . . . . . . 14<br/>3.5 Using security characteristics to test factor pricing models . . . . . . . . . . . . . . . 17<br/>3.5.1 Consistency and asymptotic normality of the CSR estimator . . . . . . . . . 19<br/>3.5.2 Misspeci&amp;#xC;cation bias and protection against spurious factors . . . . . . . . . . 20<br/>3.6 Time-varying security characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 21<br/>3.6.1 No pricing restrictions imposed on traded factors . . . . . . . . . . . . . . . . 21<br/>3.6.2 Traded factors with imposed pricing restrictions . . . . . . . . . . . . . . . . 25<br/>3.6.3 Using time-average characteristics to avoid the bias . . . . . . . . . . . . . . . 29<br/>3.7 N-consistency of the CSR estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br/>4 Maximum Likelihood Methods 36<br/>4.1 Nontraded factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br/>4.2 Some factors are traded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br/>4.3 Single risk-free lending and borrowing rates with portfolio returns as factors . . . . . 39<br/>5 The Generalized Method of Moments 40<br/>5.1 An overview of the GMM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41<br/>5.2 Evaluating beta pricing models using the beta representation . . . . . . . . . . . . . 42<br/>5.3 Evaluating beta pricing models using the stochastic discount factor representation . 46<br/>5.4 Models with time-varying betas and risk premia . . . . . . . . . . . . . . . . . . . . . 50<br/>6 Conclusions 56<br/>1 Introduction<br/>Financial assets exhibit wide variation in their historical average returns. For example, during the<br/>period from 1926 to 1999 large stocks earned an annualized average return of 13.0%, whereas longterm<br/>bonds earned only 5.6%. Small stocks earned 18.9% - substantially higher than large stocks.<br/>These di&amp;#xB;erences are statistically and economically signi&amp;#xC;cant (Jagannathan and McGrattan (1996),<br/>Ferson and Jagannathan (1996)). Furthermore, such signi&amp;#xC;cant di&amp;#xB;erences in average returns are<br/>also observed among other classes of stocks. If investors were rational, they would have anticipated<br/>such di&amp;#xB;erences. Nevertheless, they still preferred to hold &amp;#xC;nancial assets with such widely di&amp;#xB;erent<br/>expected returns. A natural question that arises is why this is the case. A variety of asset pricing<br/>models have been proposed in the literature for understanding why di&amp;#xB;erent assets earn di&amp;#xB;erent<br/>expected rates of return. According to these models di&amp;#xB;erence assets earn di&amp;#xB;erent expected returns<br/>only because they di&amp;#xB;er in their systematic risk. The models di&amp;#xB;er based on the stand they take<br/>regarding what constitutes systematic risk. Among them, the linear beta pricing models form an<br/>important class.<br/>According to linear beta pricing models, a few economy-wide pervasive factors are su&amp;#xE;cient to<br/>represent systematic risk, and the expected return on an asset is a linear function of its factor betas<br/>(Ross (1976), Connor (1984)). Some beta pricing models specify what the risk factor should be<br/>based on theoretical arguments. According to the standard Capital Asset Pricing Model (CAPM)<br/>of Sharpe (1964) and Lintner (1965), the return on the market portfolio of all assets that are in<br/>positive net supply in the economy is the relevant risk factor. Other models specify factors based<br/>on economic intuition and introspection. For example, Chen, Roll and Ross (1986) specify unanticipated<br/>changes in the term premium, default premium, the growth rate of industrial production<br/>and in<br/>ation as the factors, whereas Fama and French (1993) construct factors that capture the<br/>size and book-to-market e&amp;#xB;ects documented in the literature and examine if these are su&amp;#xE;cient to<br/>capture all economy-wide pervasive sources of risk. Campbell (1996) and Jagannathan and Wang<br/>(1996) use innovations to labor income as an aggregate risk factor. Another approach is to identify<br/>the pervasive risk factors based on systematic statistical analysis of historical return data as in<br/>Connor and Korajczyk (1988) and Lehmann and Modest (1988).<br/>In this chapter we discuss econometric methods that have been used to evaluate linear beta<br/>pricing models using historical return data on a large cross section of stocks. Three approaches<br/>have been suggested in the literature for examining linear beta pricing models: (a) Cross sectional<br/>regressions method; (b) Maximum Likelihood (ML) methods; and (c) Generalized method of moments<br/>(GMM). Shanken (1992) and MacKinlay and Richardson (1991) show that the cross-sectional<br/>method is asymptotically equivalent to ML and GMM when returns are conditionally homoscedastic.<br/>In view of this, we focus our attention primarily on the cross-sectional regression method since<br/>1<br/>it is more robust and easier to implement in large cross sections, and provide only a brief overview<br/>of the use of ML and GMM.<br/>Fama and MacBeth (1973) developed the two pass cross sectional regression method to examine<br/>whether the relation between expected return and factor betas are linear. Betas are estimated using<br/>time series regression in the &amp;#xC;rst pass and the relation between returns and betas are estimated<br/>using a second pass cross sectional regression. The use of estimated betas in the second pass<br/>introduces the classical errors-in-variables problem. The standard method for handling errors in<br/>variables problem is to group stocks into portfolios following Black, Jensen and Scholes (1972).<br/>Since each portfolio has a large number of individual stocks, portfolio betas are estimated with<br/>su&amp;#xE;cient precision and this fact allows one to ignore the errors-in-variables problem as being of<br/>second order in importance. One, however, has to be careful to ensure that the portfolio formation<br/>method does not highlight or mask characteristics in the data that have valuable information about<br/>the validity of the asset pricing model under examination. Put in other words, one has to avoid<br/>data snooping biases discussed in Lo and MacKinlay (1990).<br/>Shanken (1992) provided the &amp;#xC;rst comprehensive analysis of the statistical properties of the<br/>classical two-pass estimator under the assumption that returns and factors exhibit conditional homoscedasticity.<br/>He demonstrated ed how to take into account the sampling errors in the betas<br/>estimated in the &amp;#xC;rst pass and generalized-least-squares in the second stage cross-sectional regressions.<br/>Given these adjustments, Shanken (1992) conjectured that it may not be necessary to group<br/>securities into portfolios in order to address the errors in variables problem. Brennan, Chordia and<br/>Subrahmanyam (1998) make the interesting observation that the errors in variables problem can<br/>be avoided without grouping securities into portfolios by using risk-adjusted returns as dependent<br/>variables in tests of linear beta pricing models, provided all the factors are excess returns on traded<br/>assets. However, the relative merits of this approach as compared to portfolio grouping procedures<br/>has not been examined in the literature.<br/>Jagannathan andWang (1998) extended Shanken's analysis to allow for conditional heteroscedasticity<br/>and consider the case where the model is misspeci&amp;#xC;ed. This may happen even when the model<br/>holds in the population, if the econometrician uses the wrong factors or misses factors in computing<br/>factor betas. When the linear factor pricing model is correctly speci&amp;#xC;ed, &amp;#xC;rm characteristics such<br/>as &amp;#xC;rm size should not be able to explain expected return variations in the cross section of stocks.<br/>In the case of misspeci&amp;#xC;ed factor models, Jagannathan and Wang (1998) showed that the t-values<br/>associated with &amp;#xC;rm characteristics will typically be large. Hence, model misspeci&amp;#xC;cation can be<br/>detected using &amp;#xC;rm characteristics in cross-sectional regression. Such a test does not require that<br/>the number of assets be small relative to the length of the time series of observations on asset<br/>returns, as is the case with standard multivariate tests of linearity.<br/>2<br/>Gibbons (1982) showed that the classical maximum likelihood method can be used to estimate<br/>and test linear beta pricing models when stock returns are i.i.d and jointly normal. Kandel (1984)<br/>developed a straight forward computational procedure for implementing the maximum likelihood<br/>method. Shanken (1992) extended it further and showed that the cross sectional regression approach<br/>can be made asymptotically as e&amp;#xE;cient as the maximum likelihood method. Kim (1985) developed<br/>a maximum likelihood procedure that allows for the use of betas estimated using past data. Jobson<br/>and Korkie (1982) and MacKinlay (1987) developed exact multivariate tests for the CAPM and<br/>Gibbons, Ross and Shanken (1989) exact multivariate tests for linear beta pricing models when<br/>there is a risk free asset.<br/>MacKinlay and Richardson (1991) show how to estimate the parameters of the CAPM by<br/>applying the GMM to its beta representation. They illustrate the bias in the tests based on<br/>standard maximum likelihood methods when stock returns exhibit contemporaneous conditional<br/>heteroscedasticity and show that the GMM estimator and the maximum likelihood method are<br/>equivalent under conditional homoscedasticity. An advantage of using the GMM is that it allows<br/>estimation of model parameters in a single pass thereby avoiding the error-in-variables problem.<br/>Linear factor pricing models can also be estimated by applying the GMM to their stochastic discount<br/>factor (SDF) representation. Jagannathan and Wang (2001) show that parameters estimated by<br/>applying the GMM to the SDF representation and the beta representation of linear beta pricing<br/>models are asymptotically equivalent.<br/>The rest of the chapter is organized as follows. In Section 2 we set up the necessary notation and<br/>describe the general linear beta pricing model. We discuss in detail the two pass cross sectional<br/>regression method in Section 3 and provide an overview of the maximum likelihood methods in<br/>Section 4 and the GMM in Section 5. We summarize in Section 6.2008-07-22 21:20:47 http://www.netyi.net/training/6158dab7-942d-4e04-ba83-6759dbb19219Portfolio Choice Problems&#164;<br/>Michael W. Brandt<br/>Fuqua School of Business<br/>Duke Universityy<br/>and NBER<br/>August 2004<br/>&#164;To appear in Y. A?3t-Sahalia and L.P. Hansen, eds., Handbook of Financial Econometrics, Elsevier Science:<br/>Amsterdam. I thank Pedro Santa-Clara and Luis Viceira for their comments and suggestions.<br/>yDurham, NC 27708-1020. Phone: (919) 660-1948. E-mail:mbrandt@duke.edu.<br/>1 Introduction<br/>After years of relative neglect in academic circles, portfolio choice problems are again at<br/>the forefront of ˉnancial research. The economic theory underlying an investor's optimal<br/>portfolio choice, pioneered by Markowitz (1952), Merton (1969,1971), Samuelson (1969), and<br/>Fama (1970), is by now well understood. The renewed interest in portfolio choice problems<br/>follows the relatively recent empirical evidence of time-varying return distributions (e.g.,<br/>predictability and conditional heteroskedasticity) and is fueled by realistic issues including<br/>model and parameter uncertainty, learning, background risks, and frictions. The general<br/>focus of the current academic research is to identify key aspects of real-world portfolio choice<br/>problems and to understand qualitatively as well as quantitatively their role in the optimal<br/>investment decisions of individuals and institutions.<br/>Whether for academic researchers studying the portfolio choice implications of return<br/>predictability, for example, or for practitioners whose livelihood depends on the outcome of<br/>their investment decisions, a critical step in solving realistic portfolio choice problems is to<br/>relate the theoretical formulation of the problem and its solution to the data. There are a<br/>number of ways to accomplish this task, ranging from calibration with only vague regard for<br/>the data to decision theoretic approaches which explicitly incorporate the speciˉcation of<br/>the return model and the associated statistical inferences in the investor's decision process.<br/>Surprisingly, given the practical importance of portfolio choice problems, no single econo-<br/>metric approach has emerged yet as clear favorite. Since each approach has its advantages<br/>and disadvantages, an approach favored in one context is often less attractive in another.<br/>This chapter is devoted to the econometric treatment of portfolio choice problems. The<br/>goal is to describe, discuss, and illustrate through examples the di?erent econometric ap-<br/>proaches proposed in the literature for relating the theoretical formulation and solution of a<br/>portfolio choice problem to the data. The chapter is intended for academic researchers who<br/>seek an introduction to the empirical implementation of portfolio choice problems as well as<br/>for practitioners as a review of the academic literature on the topic.<br/>The chapter is divided into three parts. Section 2 reviews the theory of portfolio choice in<br/>discrete and continuous time. It also discusses a number of modeling issues and extensions<br/>that arise in formulating the problem. Section 3 presents the two traditional econometric<br/>approaches to portfolio choice problems: plug-in estimation and Bayesian decision theory.<br/>In Section 4, I then describe a more recently developed econometric approach for drawing<br/>inferences about optimal portfolio weights without modeling return distributions.2008-07-22 21:19:06 http://www.netyi.net/training/771985ad-c84f-4589-a00f-823a621d68fcParametric and Nonparametric<br/>Volatility Measurement*<br/>Torben G. Andersena, Tim Bollerslevb, and Francis X. Dieboldc<br/>July 2002<br/>__________________<br/>* This paper is prepared for Yacine A?t-Sahalia and Lars Peter Hansen (eds.), Handbook of Financial Econometrics,<br/>Amsterdam: North Holland. We are grateful to the National Science Foundation for research support, and to Nour<br/>Meddahi, Neil Shephard and Sean Campbell for useful discussions and detailed comments on earlier drafts.<br/>a Department of Finance, Kellogg School of Management, Northwestern University, Evanston, IL 60208, and NBER<br/>phone: 847-467-1285, e-mail: t-andersen@kellogg.northwestern.edu<br/>b Departments of Economics and Finance, Duke University, Durham, NC 27708, and NBER<br/>phone: 919-660-1846, e-mail: boller@econ.duke.edu<br/>c Departments of Economics, Finance and Statistics, University of Pennsylvania, Philadelphia, PA 19104, and NBER<br/>phone: 215-898-1507, e-mail: fdiebold@sas.upenn.edu<br/>Table of Contents<br/>Abstract<br/>1. Introduction<br/>2. Volatility Definitions<br/>2.1. Continuous-Time No-Arbitrage Pricing<br/>2.2. Notional, Expected, and Instantaneous Volatility<br/>2.3. Volatility Models and Measurements<br/>3. Parametric Methods<br/>3.1. Discrete-Time Models<br/>3.1.1. ARCH Models<br/>3.1.2. Stochastic Volatility Models<br/>3.2. Continuous-Time Models<br/>3.2.1. Continuous Sample Path Diffusions<br/>3.2.1. Jump Diffusions and L&#233;vy Driven Processes<br/>4. Nonparametric Methods<br/>4.1. ARCH Filters and Smoothers<br/>4.2. Realized Volatility<br/>5. Conclusion<br/>References<br/>ABSTRACT<br/>Volatility has been one of the most active areas of research in empirical finance and time series<br/>econometrics during the past decade. This chapter provides a unified continuous-time,<br/>frictionless, no-arbitrage framework for systematically categorizing the various volatility<br/>concepts, measurement procedures, and modeling procedures. We define three different<br/>volatility concepts: (i) the notional volatility corresponding to the ex-post sample-path return<br/>variability over a fixed time interval, (ii) the ex-ante expected volatility over a fixed time<br/>interval, and (iii) the instantaneous volatility corresponding to the strength of the volatility<br/>process at a point in time. The parametric procedures rely on explicit functional form<br/>assumptions regarding the expected and/or instantaneous volatility. In the discrete-time<br/>ARCH class of models, the expectations are formulated in terms of directly observable<br/>variables, while the discrete- and continuous-time stochastic volatility models involve latent<br/>state variable(s). The nonparametric procedures are generally free from such functional form<br/>assumptions and hence afford estimates of notional volatility that are flexible yet consistent<br/>(as the sampling frequency of the underlying returns increases). The nonparametric<br/>procedures include ARCH filters and smoothers designed to measure the volatility over<br/>infinitesimally short horizons, as well as the recently-popularized realized volatility measures<br/>for (non-trivial) fixed-length time intervals.<br/>1 See, for example, Bollerslev, Chou and Kroner (1992).<br/>-1-<br/>1. INTRODUCTION<br/>Since Engle’s (1982) seminal paper on ARCH models, the econometrics literature has focused<br/>considerable attention on time-varying volatility and the development of new tools for<br/>volatility measurement, modeling and forecasting. These advances have in large part been<br/>motivated by the empirical observation that financial asset return volatility is time-varying in a<br/>persistent fashion, across assets, asset classes, time periods, and countries.1 Asset return<br/>volatility, moreover, is central to finance, whether in asset pricing, portfolio allocation, or risk<br/>management, and standard financial econometric methods and models take on a very different,<br/>conditional, flavor when volatility is properly recognized to be time-varying.<br/>The combination of powerful methodological advances and tremendous relevance in empirical<br/>finance produced explosive growth in the financial econometrics of volatility dynamics, with<br/>the econometrics and finance literatures cross-fertilizing each other furiously. Initial<br/>developments were tightly parametric, but the recent literature has moved in less parametric,<br/>and even fully nonparametric, directions. Here we review and provide a unified framework for<br/>interpreting both the parametric and nonparametric approaches.<br/>In section 2, we define three different volatility concepts: (i) the notional volatility<br/>corresponding to the ex-post sample-path return variability over a fixed time interval, (ii) the<br/>ex-ante expected volatility over a fixed time interval, and (iii) the instantaneous volatility<br/>corresponding to the strength of the volatility process at a point in time.<br/>In section 3, we survey parametric approaches to volatility modeling, which are based on<br/>explicit functional form assumptions regarding the expected and/or instantaneous volatility. In<br/>the discrete-time ARCH class of models, the expectations are formulated in terms of directly<br/>observable variables, while the discrete- and continuous-time stochastic volatility models both<br/>involve latent state variable(s).<br/>In section 4, we survey nonparametric approaches to volatility modeling, which are generally<br/>free from such functional form assumptions and hence afford estimates of notional volatility<br/>that are flexible yet consistent (as the sampling frequency of the underlying returns increases).<br/>The nonparametric approaches include ARCH filters and smoothers designed to measure the<br/>volatility over infinitesimally short horizons, as well as the recently-popularized realized<br/>volatility measures for (non-trivial) fixed-length time intervals.<br/>We conclude in section 5 by highlighting promising directions for future research.2008-07-22 21:17:46 http://www.netyi.net/training/3a64a970-0616-4bb1-ab01-1a1ce9f86739Operator Methods for Continuous-Time Markov Processes&#164;<br/>Yacine A?3t-Sahalia<br/>Department of Economics<br/>Princeton University<br/>Lars Peter Hansen<br/>Department of Economics<br/>The University of Chicago<br/>Jos?e A. Scheinkman<br/>Department of Economics<br/>Princeton University<br/>First Draft: November 2001. This Version: August 21, 2004<br/>1 Introduction<br/>Our chapter surveys a set of mathematical and statistical tools that are valuable in understanding and charac-<br/>terizing nonlinear Markov processes. Such processes are used extensively as building blocks in economics and<br/>ˉnance. In these literatures, typically the local evolution or short-run transition is speciˉed. We concentrate<br/>on the continuous limit in which case it is the instantaneous transition that is speciˉed. In understanding<br/>the implications of such a modelling approach we show how to infer the intermediate and long-run properties<br/>from the short-run dynamics. To accomplish this we describe operator methods and their use in conjunction<br/>with continuous-time stochastic process models.<br/>Operator methods begin with a local characterization of the Markov process dynamics. This local speciˉ-<br/>cation takes the form of an inˉnitesimal generator. The inˉnitesimal generator is itself an operator mapping<br/>test functions into other functions. From the inˉnitesimal generator, we construct a family (semigroup) of<br/>conditional expectation operators. The operators exploit the time-invariant Markov structure. Each operator<br/>in this family is indexed by the forecast horizon, the interval of time between the information set used for<br/>prediction and the object that is being predicted. Operator methods allow us to ascertain global, and in par-<br/>ticular, long-run implications from the local or inˉnitesimal evolution. These global implications are re&#176;ected<br/>in a) the implied stationary distribution b) the analysis of the eigenfunctions of the generator that dominate<br/>in the long run, c) the construction of likelihood expansions and other estimating equations.<br/>The methods we describe in this chapter are designed to show how global and long-run implications follow<br/>from local characterizations of the time series evolution. This connection between local and global properties<br/>is particularly challenging for nonlinear time series models. In spite of this complexity, the Markov structure<br/>makes characterizations of the dynamic evolution tractable. In addition to facilitating the study of a given<br/>Markov process, operator methods provide characterizations of the observable implications of potentially rich<br/>families of such processes. These methods can be incorporated into statistical estimation and testing. While<br/>many Markov processes used in practice are formally misspeciˉcied, operator methods are useful in exploring<br/>the speciˉc nature and consequences of this misspeciˉcation.<br/>&#164;All three authors gratefully acknowledge ˉnancial support from the National Science Foundation.<br/>1<br/>Section 2 describes the underlying mathematical methods and notation. Section 3 studies Markov models<br/>through their implied stationary distributions. Section 4 gives some operator characterizations and related<br/>expansions used to characterize transition dynamics. Section 7 describes the properties of some parameter<br/>estimators. Section 6 investigates alternative ways to characterize the observable implications of various<br/>Markov models, and the tests that can be constructed based on those characterizations.2008-07-22 21:16:34 http://www.netyi.net/training/cc301fc0-9f4f-4d9e-ac92-2fd81e0bb5beMeasuring and Modeling Variation in the Risk-Return<br/>TradeoT ￡<br/>Martin Lettau<br/>New York University, CEPR, NBER<br/>Sydney C. Ludvigson<br/>New York University and NBER<br/>Preliminary<br/>Comments Welcome<br/>First draft: July 24, 2001<br/>This draft: December 3, 2003<br/>￡Lettau: Department of Finance, Stern School of Business, New York University, 44 West Fourth Street,<br/>New York, NY 10012-1126; Email: mlettau@stern.nyu.edu, Tel: (212) 998-0378; Fax: (212) 995-4233.<br/>Ludvigson: Department of Economics, New York University, 269 Mercer Street, 7th Floor, New York,<br/>NY 10003; Email: sydney.ludvigson@nyu.edu; Tel: (212) 998-8927; Fax: (212) 995-4186.<br/>This paper has been prepared for the Handbook of Financial Econometrics, edited by Yacine Ait-<br/>Sahalia and Lars Peter Hansen. Updated versions, along with the data on dcayt, may be found at<br/>http://www.stern.nyu.edu/&#243; mlettau and http://www.econ.nyu.edu/user/ludvigsons/. Lettau ac-<br/>knowledges ?nancial support from the National Science Foundation; Ludvigson acknowledges ?nancial sup-<br/>port from the Alfred P. Sloan Foundation and the National Science Foundation. We are grateful to Michael<br/>Brandt, Rob Engle, Stijn Van Nieuwerburgh, and Jessica Wachter for helpful comments. Nathan Barczi and<br/>Adam Kolasinski provided excellent research assistance. Any errors or omissions are the responsibility of<br/>the authors.<br/>Measuring and Modeling Variation in the Risk-Return TradeoT<br/>Abstract<br/>Are excess stock market returns predictable over time and, if so, at what horizons and<br/>with which economic indicators? Can stock return predictability be explained by changes<br/>in stock market volatility? How does the mean return per unit risk change over time? This<br/>chapter reviews what is known about the time-series evolution of the risk-return tradeoT for<br/>stock market investment, and presents some new empirical evidence using a proxy for the log<br/>consumption-aggregate wealth ratio as a predictor of both the mean and volatility of excess<br/>stock market returns.<br/>We characterize the risk-return tradeoT as the conditional expected excess return on a<br/>broad stock market index divided by its conditional standard deviation, a quantity commonly<br/>known as the Sharpe ratio. Our own investigation suggests that variation in the equity risk-<br/>premium is strongly negatively linked to variation in market volatility, at odds with leading<br/>asset pricing models. Since the conditional volatility and conditional mean move in opposite<br/>directions, the degree of countercyclicality in the Sharpe ratio that we document here is<br/>far more dramatic than that produced by existing equilibrium models of ?nancial market<br/>behavior, which completely miss the sheer magnitude of variation in the price of stock market<br/>risk; leading asset pricing paradigms leave a ?Sharpe ratio volatility puzzle? that remains to<br/>be explained.<br/>JEL: G10, G12.<br/>1 Introduction<br/>Financial markets are often hard to understand. Stock prices frequently seem volatile and<br/>unpredictable, and researchers have devoted signi?cant resources to understanding the be-<br/>havior of expected returns relative to the risk of stock market investment. Are excess stock<br/>market returns predictable over time and, if so, at what horizons and with which economic<br/>indicators? Can stock return predictability be explained by changes in stock market volatil-<br/>ity? How does the mean return per unit risk change over time? For academic researchers,<br/>the progression of empirical evidence aimed at these questions has presented a continuing<br/>challenge to asset pricing theory and an important road map for future inquiry. For many<br/>investment professionals, ?nding practical answers to these questions is the fundamental<br/>purpose of ?nancial economics, as well as its principal reward.<br/>Despite both the theoretical and practical importance of these issues, relatively little is<br/>known about how the risk-return tradeoT varies over the business cycle or with key macroe-<br/>conomic indicators. This chapter reviews the state of knowledge on such variation for stock<br/>market investment, and presents some new empirical evidence based on information con-<br/>tained in aggregate consumption and aggregate labor income. We de?ne the risk-return<br/>tradeoT as the conditional expected excess return on a broad stock market index divided by<br/>its conditional standard deviation, a quantity commonly known as the Sharpe ratio. Our<br/>study focuses not on the unconditional value of this ratio, but on its evolution through time.<br/>Understanding the time-series properties of the Sharpe ratio is crucial to the development<br/>of theoretical models capable of explaining observed patterns of stock market predictability<br/>and volatility. For example, Hansen and Jagannathan (1991) showed that the maximum<br/>value of the Sharpe ratio places restrictions on the volatility of the set of discount factors<br/>that can be used to price returns. The same reasoning implies that the pattern of time-<br/>series variation in the Sharpe ratio will also place restrictions on the set of discount factors<br/>capable of pricing equity returns. In addition, the behavior of the Sharpe ratio over time<br/>is fundamental for assessing whether stocks are safer in the long run than they are in the<br/>short run, as increasingly advocated by popular guides to investment strategy (e.g., Siegel<br/>(1998)). Only if the Sharpe ratio grows more quickly than the square root of the horizon?so<br/>that the variance of the return grows more slowly than its mean?are stocks safer investments<br/>in the long run than they are in the short run. Such a dynamic pattern is not possible if<br/>stock returns are unpredictable, i.i.d. random variables. Thus, understanding the time-series<br/>behavior of the Sharpe ratio not only provides a benchmark for theoretical progress, it has<br/>3<br/>profound implications for investment professionals concerned with strategic asset allocation.<br/>The two components of the risk-return relation (the numerator and the denominator of<br/>the Sharpe ratio) are the conditional mean excess stock return, and the conditional standard<br/>deviation of the excess return. We focus here on empirically measuring and statistically<br/>modeling each of these components separately, a process that can be uni?ed to reveal an<br/>estimate of the conditional Sharpe ratio, or price of stock market risk. Section 2 discusses<br/>estimation of the conditional mean of excess stock returns. In this section we evaluate the<br/>statistical evidence for stock return predictability and review the range of indicators with<br/>which such predictability has been associated. Taken together, this evidence suggests that<br/>excess returns on broad stock market indexes are predictable at long-horizons, implying that<br/>the reward for bearing risk varies over time.<br/>One possible explanation for time-variation in the equity risk premium is time variation<br/>in stock market volatility. Section 3 reviews the evidence for time-variation in stock market<br/>volatility. In many classic asset pricing models, the equity risk premium varies proportionally<br/>with stock market volatility. These models require that periods of high excess stock returns<br/>coincide with periods of high stock market volatility, implying a constant price of risk. It<br/>follows that variation in the equity risk premium must be perfectly positively correlated with<br/>variation in stock market volatility.<br/>The important empirical question is whether such a positive correlation between the<br/>mean and volatility of returns exists, implying a constant Sharpe ratio. Section 4 ties the<br/>evidence on the conditional mean of excess returns in with that on the conditional variance<br/>to derive implications for the time-series behavior of the conditional Sharpe ratio. Existing<br/>empirical evidence on the sign of the relationship between the conditional mean and the<br/>conditional volatility of excess stock returns is mixed and somewhat weak. This may be<br/>because some studies have relied on parametric or semi-parametric ARCH-like models of<br/>volatility that impose a relatively high degree of structure about which there is little direct<br/>empirical evidence. Others have used predictive variables for volatility that are only weakly<br/>related to the ?rst moments of returns, and vice versa. Finally, it has been di&#225;cult to explain<br/>high risk premia with high volatility because evidence suggests that returns are predicable<br/>at quarterly and longer horizons, while variation in stock market volatility has, to date, been<br/>most evident in high frequency (e.g., daily) data.<br/>In addition to reviewing existing evidence, this chapter presents some new evidence on<br/>the risk-return tradeoT. We ?nd that a proxy for the log consumption-aggregate wealth ratio,<br/>a variable shown elsewhere to predict excess returns and constructed using information on<br/>4<br/>aggregate consumption and labor income, is also a strong predictor of stock market volatility.<br/>These ?ndings diTer from existing evidence because they reveal the presence of at least one<br/>observable conditioning variable that strongly forecasts both the mean and volatility of<br/>returns. Moreover, these results show that the evidence for changing stock market risk is<br/>not con?ned to high frequency data: stock market volatility is forecastable over horizons<br/>ranging from one quarter to six years.<br/>These ?ndings imply that movements in the equity risk-premium are linked empirically<br/>to stock market volatility. In addition, the predictability patterns we ?nd for excess returns<br/>and volatility imply pronounced countercyclical variation in the Sharpe ratio. This evidence<br/>weighs against many time-honored asset pricing models that specify a constant price of risk<br/>(for example, the static capital asset pricing model (CAPM) of Sharpe (1964) and Lintner<br/>(1965)), and toward more recent paradigms capable of rationalizing a countercyclical Sharpe<br/>ratio (e.g., Campbell and Cochrane (1999); Barberis, Huang, and Santos (2001)). Although<br/>these more recent frameworks imply a positive correlation between the conditional mean<br/>and conditional volatility of returns, unlike the classic asset pricing models this positive<br/>correlation is not perfect, making countercyclical variation in the Sharpe ratio possible.<br/>Yet despite evidence that the Sharpe ratio varies countercyclically, our results neverthe-<br/>less present a problem for modern-day asset pricing theory. Instead of ?nding a positive<br/>correlation between the condition mean and conditional volatility, we ?nd a strong negative<br/>correlation, consistent with the ?ndings of a number of other studies discussed below. More<br/>signi?cantly, since the conditional volatility and conditional mean move in opposite direc-<br/>tions, the magnitude of countercyclicality in the Sharpe ratio that we document here is far<br/>more dramatic than that produced by leading asset pricing models capable of generating a<br/>countercyclical price of risk. These results suggest that predictability of excess stock returns<br/>cannot be readily explained by changes in stock market volatility.<br/>Even if stock market volatility were constant, predictable variation in excess stock re-<br/>turns might be explained by time variation in consumption volatility. In a wide-range of<br/>equilibrium asset pricing models, more risky consumption streams require asset markets<br/>that, in equilibrium, deliver a higher mean return per unit risk. Some variation in aggregate<br/>consumption volatility is evident in the data, as we document here. However, this varia-<br/>tion is small and we conclude that changes in consumption risk, as measured by changes in<br/>the volatility of consumption growth, are insu&#225;ciently important empirically to explain the<br/>extreme swings in the Sharpe ratio that we ?nd here.<br/>2008-07-22 21:15:30 http://www.netyi.net/training/0248ef89-a9f3-4731-8785-900594bf7696MCMC Methods for Continuous-Time<br/>Financial Econometrics<br/>Michael Johannes and Nicholas Polson?<br/>December 22, 2003<br/>Abstract<br/>This chapter develops Markov Chain Monte Carlo (MCMC) methods for Bayesian<br/>inference in continuous-time asset pricing models. The Bayesian solution to the inference<br/>problem is the distribution of parameters and latent variables conditional on observed<br/>data, and MCMC methods provide a tool for exploring these high-dimensional,<br/>complex distributions. We first provide a description of the foundations and mechanics<br/>of MCMC algorithms. This includes a discussion of the Clifford-Hammersley<br/>theorem, the Gibbs sampler, the Metropolis-Hastings algorithm, and theoretical convergence<br/>properties of MCMC algorithms. We next provide a tutorial on building<br/>MCMC algorithms for a range of continuous-time asset pricing models. We include<br/>detailed examples for equity price models, option pricing models, term structure models,<br/>and regime-switching models. Finally, we discuss the issue of sequential Bayesian<br/>inference, both for parameters and state variables.<br/>?We would especially like to thank Chris Sims and the editors, Yacine Ait-Sahalia and Lars Hansen. We<br/>also thank Mark Broadie, Mike Chernov, Anne Gron, Paul Glasserman, and Eric Jacquier for their helpful<br/>comments. Johannes is at the Graduate School of Business, Columbia University, 3022 Broadway, NY,<br/>NY, 10027, mj335@columbia.edu. Polson is at the Graduate School of Business, University of Chicago,<br/>1101 East 58th Street, Chicago IL 60637, ngp@gsb.uchicago.edu.<br/>1<br/>Contents<br/>1 Introduction 4<br/>2 Overview of Bayesian Inference and MCMC 7<br/>2.1 MCMC Simulation and Estimation . . . . . . . . . . . . . . . . . . . . . . 8<br/>2.2 Bayesian Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9<br/>3 MCMC: Methods and Theory 12<br/>3.1 Clifford-Hammersley Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 12<br/>3.2 Gibbs Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14<br/>3.3 Metropolis-Hastings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15<br/>3.4 Convergence Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19<br/>3.4.1 Convergence ofMarkov Chains . . . . . . . . . . . . . . . . . . . . 20<br/>3.4.2 Convergence ofMCMC algorithms . . . . . . . . . . . . . . . . . . 21<br/>3.5 MCMCAlgorithms: Issues and Practical Recommendations . . . . . . . . 26<br/>4 Bayesian Inference and Asset Pricing Models 30<br/>4.1 States Variables and Prices . . . . . . . . . . . . . . . . . . . . . . . . . . . 31<br/>4.2 Time-discretization: computing p (Y |X, Θ) and p (X|Θ) . . . . . . . . . . . 34<br/>4.3 Parameter Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br/>5 Asset Pricing Applications 39<br/>5.1 EquityAsset PricingModels . . . . . . . . . . . . . . . . . . . . . . . . . . 39<br/>5.1.1 Geometric BrownianMotion . . . . . . . . . . . . . . . . . . . . . . 39<br/>5.1.2 Black-Scholes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41<br/>5.1.3 AMultivariateVersion ofMerton’sModel . . . . . . . . . . . . . . 43<br/>5.1.4 Time-Varying Equity Premium . . . . . . . . . . . . . . . . . . . . 47<br/>5.1.5 Log-Stochastic VolatilityModels . . . . . . . . . . . . . . . . . . . . 53<br/>5.1.6 Alternative Stochastic VolatilityModels . . . . . . . . . . . . . . . 58<br/>5.2 Term Structure Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63<br/>5.2.1 Vasicek’s Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63<br/>5.2.2 Vasicek with Jumps . . . . . . . . . . . . . . . . . . . . . . . . . . . 66<br/>5.2.3 The CIR model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70<br/>5.3 Regime Switching Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 72<br/>2<br/>6 Sequential Inference: Filtering 74<br/>6.1 The Particle Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75<br/>6.1.1 Adapting the particle filter to continuous-timemodels . . . . . . . . 79<br/>6.2 Practical Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81<br/>7 Conclusions and Future Directions 83<br/>8 References 85<br/>3<br/>1 Introduction<br/>Dynamic asset pricing theory uses arbitrage and equilibrium arguments to derive the functional<br/>relationship between asset prices and the fundamentals of the economy: state variables,<br/>structural parameters and market prices of risk. Continuous-time models are the<br/>centerpiece of this approach due to their analytical tractability. In many cases, these models<br/>lead to closed form solutions or easy to solve differential equations for objects of interest<br/>such as prices or optimal portfolio weights. The models are also appealing from an empirical<br/>perspective: through a judicious choice of the drift, diffusion, jump intensity and jump<br/>distribution, these models accommodate a wide range of dynamics for state variables and<br/>prices.<br/>Empirical analysis of dynamic asset pricing models tackles the inverse problem: extracting<br/>information about latent state variables, structural parameters and market prices<br/>of risk from observed prices. The Bayesian solution to the inference problem is the distribution<br/>of the parameters, Θ, and state variables, X, conditional on observed prices, Y. This<br/>posterior distribution, p (Θ, X|Y ), combines the information in the model and the observed<br/>prices and is the key to inference on parameters and state variables.<br/>This chapter describes Markov Chain Monte Carlo (MCMC) methods for exploring<br/>the posterior distributions generated by continuous-time asset pricing models. MCMC<br/>samples from these high-dimensional, complex distributions by generating a Markov Chain<br/>over (Θ, X), ?Θ(g), X(g)aG<br/>g=1, whose equilibrium distribution is p (Θ, X|Y ). The Monte<br/>Carlo method uses these samples for numerical integration for parameter estimation, state<br/>estimation and model comparison.<br/>Characterizing p (Θ, X|Y ) in continuous-time asset pricing models is difficult for a variety<br/>of reasons. First, prices are observed discretely while the theoretical models specify<br/>that prices and state variables evolve continuously in time. Second, in many cases, the<br/>state variables are latent from the researcher’s perspective. Third, p (Θ, X|Y ) is typically<br/>of very high dimension and thus standard sampling methods commonly fail. Fourth, many<br/>continuous-time models of interest generate transition distributions for prices and state variables<br/>that are non-normal and non-standard, complicating standard estimation methods<br/>such as MLE or GMM. Finally, in term structure and option pricing models, parameters<br/>enter nonlinearly or even in a non-analytic form as the implicit solution to ordinary or<br/>partial differential equations. We show that MCMC methods tackle all of these issues.<br/>To frame the issues involved, it is useful to consider the following example: Suppose on<br/>4<br/>(Ω,F, P) an asset price, St, and its stochastic variance, Vt, jointly solve:<br/>dSt = St (rt + μt) dt + StpVtdW s<br/>t (P) + dμXNt(P)<br/>j=1<br/>Sτ j? ?eZj(P) ? 1￠?? μPt<br/>Stdt (1)<br/>dVt = κv (θv ? Vt) dt + σvpVtdW v<br/>t (P) (2)<br/>where W s<br/>t (P) and W v<br/>t (P) are Brownian motions, Nt (P) counts the number of jump times,<br/>τ j , prior to time t, μt is the equity risk premium, μPt<br/>St is the jump compensator, Zj (P) are<br/>the jump sizes, and rt is the spot interest rate. Researchers also often observe derivative<br/>prices, such as options. To price these derivatives, asset pricing theory asserts the existence<br/>of a probability measure, Q, such that<br/>dSt = rtStdt + StpVtdW s<br/>t (Q) + dμXNt(Q)<br/>j=1<br/>Sτ j? ?eZj(Q) ? 1￠?? μQt<br/>Stdt<br/>dVt = ￡κv (θv ? Vt) + λQv<br/>Vt&#164;dt + σvpVtdW v<br/>t (Q)<br/>where all random variables are now defined on (Ω,F,Q). Here λQv<br/>is the diffusive “price of<br/>volatility risk,” and μQt<br/>is the jump compensator. Under Q, the price of a call option on St<br/>maturing at time T, struck at K, is<br/>Ct = C (St, Vt, Θ) = EQ &#183;expμ? Z T<br/>t<br/>rsds?(ST ? K)+ |Vt, St, Θ? (3)<br/>where Θ = ?ΘP, ΘQ￠ are the structural and risk neutral parameters. The state variables,<br/>X, consist of the volatilities, the jump times and jump sizes.<br/>The goal of empirical asset pricing is to learn about the risk neutral and objective<br/>parameters, the state variables, namely, volatility, jump times and jump sizes, and the<br/>model specification from the observed equity returns and option prices. In the case of the<br/>parameters, the marginal posterior distribution p (Θ|Y ) characterizes the sample information<br/>about the objective and risk-neutral parameters and quantifies the estimation risk:<br/>the uncertainty inherent in estimating parameters. For the state variables, the marginal<br/>distribution, p (X|Y ), combines the model and data to provide a consistent approach for<br/>separating out the effects of jumps from stochastic volatility. This is important for empirical<br/>problems such as option pricing or portfolio applications which require volatility estimates.<br/>Classical methods are difficult to apply in this model as the parameters and volatility enter<br/>in a non-analytic manner in the option pricing formula, volatility, jump times and jump<br/>sizes are latent, and the transition density for observed prices is not known.<br/>5<br/>To design MCMC algorithms for exploring p (Θ, X|Y ), we first follow Duffie (1996) and<br/>interpret asset pricing models as state space models. This interpretation is convenient for<br/>constructing MCMC algorithms as it highlights the modular nature of asset pricing models.<br/>The observation equation is the distribution of the observed asset prices conditional on the<br/>state variables and parameters while the evolution equation consists of the dynamics of<br/>state variables conditional on the parameters. In the example above, (1) and (3) form the<br/>observation equations and (2) is the evolution equation. Viewed in this manner, all asset<br/>pricing models take the general form of nonlinear, non-Gaussian state space models.2008-07-22 21:14:00 http://www.netyi.net/training/871f2cf0-7682-4f61-a09c-172b895dfc38Revised September 2004<br/>Heterogeneity and Portfolio Choice: Theory and Evidence<br/>Stephanie Curcuru1<br/>John Heaton2<br/>Deborah Lucas3<br/>Damien Moore4<br/>Abstract<br/>In this paper, we summarize and add to the evidence on the large and systematic differences in portfolio<br/>composition across individuals with varying characteristics, and evaluate some of the theories that have been<br/>proposed in terms of their ability to account for these differences. Variation in background risk exposure --<br/>from sources such as labor and entrepreneurial income or real estate holdings, and from factors such as<br/>transactions costs, borrowing constraints, restricted pension investments and life cycle considerations – can<br/>explain some but not all aspects of the observed cross-sectional variation in portfolio holdings in a traditional<br/>utility maximizing framework. In particular, fixed costs and life cycle considerations appear necessary to<br/>explain the lack of stock market participation by young and less affluent households. Remaining challenges<br/>for quantitative theories include the apparent lack of diversification in some unconstrained individual<br/>portfolios, and non-participation in the stock market by some households with significant financial wealth.<br/>1.Graduate School of Business, University of Chicago<br/>2.Graduate School of Business, University of Chicago and NBER<br/>3.Northwestern University Kellogg School of Management and the NBER<br/>4. University of Sydney<br/>Prepared for the Handbook of Financial Econometrics.<br/>2<br/>1. Introduction<br/>Data on households’ financial behavior points to considerable heterogeneity in portfolio allocations.<br/>The majority of households hold neither common stock nor other risky financial securities. Others invest in<br/>stocks almost exclusively. The extent to which risky asset holdings are diversified also varies greatly ,<br/>ranging from exclusive reliance on diversified index funds to holdings concentrated in a few individual<br/>stocks. Employees often have significant holdings in the stock of their employers. To make sense of these<br/>observations, it is useful to look for empirical regularities in the way that households with different<br/>characteristics invest their savings, and to interpret these regularities using theories of portfolio choice that<br/>allow heterogeneity among investors. In this paper, we summarize and add to the evidence on the large and<br/>systematic differences in portfolio composition across individuals with varying characteristics, and evaluate<br/>some of the theories that have been proposed in terms of their ability to account for these differences.<br/>If heterogeneity in portfolio allocations is to be explained in a traditional utility maximizing<br/>framework, it must be accounted for by heterogeneity in preferences, heterogeneity in circumstances, or a<br/>combination of the two. The recent literature on portfolio choice has emphasized both of these possibilities.<br/>Heterogeneity in circumstances encompasses a wide range of potential explanatory factors including the<br/>presence of non-diversifiable background risks, demographics, information asymmetries and transaction<br/>costs. Potential sources of non-diversifiable background risks include labor income and proprietary business<br/>income (or more broadly human capital), restricted pension investments, and owner-occupied real estate.<br/>Demographic factors include age, occupation, inherited wealth and education. Transactions costs include<br/>taxes, the fixed and variable costs of trading in securities markets, and also the time or psychic costs of<br/>learning about asset markets.<br/>To provide context for the empirical findings, we briefly review some of the extensive theoretical<br/>literature on portfolio choice, with an emphasis on calibrated models explicitly designed to quantitatively<br/>explain heterogeneity. Early papers by Mossin (1968), Merton (1969), and Samuelson (1969, 1970) were<br/>the first to address the dynamic portfolio choice problem in preference based theories. Friend and Blume<br/>3<br/>(1975) found that, consistent with average asset holdings, calibrated versions of these theoretical portfolio<br/>share rules imply a fairly even division of wealth between stocks and bonds, assuming moderate risk<br/>aversion. For the next several decades portfolio choice was thought to be a largely solved problem.<br/>Recently, however, there has been renewed interest in this area. This can be attributed to the greater<br/>availability of data that reveals apparent idiosyncrasies in individual behavior , and also to the increasing<br/>interest in the implications of incomplete markets. Both naturally lead to an examination of optimal<br/>portfolio behavior in the presence of market frictions that can affect portfolio allocation rules.1<br/>Apart from explaining data, understanding portfolio choice can shed light on a variety of broader<br/>issues. For one, asset pricing models presume a theory of portfolio choice. Unresolved issues in the asset<br/>pricing literature, such as the equity premium puzzle (Mehra and Prescott (1985)), are related to unsettled<br/>questions in the portfolio choice literature, such as explaining non-participation in the stock market, or that<br/>many portfolios are skewed towards safe assets. For instance, without the existence of a significant<br/>differential between the average return on stocks and that on short-term risk-free bonds (the equity<br/>premium), it would be quite easy to account for low stock market participation by appealing to moderate<br/>transaction costs or background risk using conventional models. Research on portfolio choice may help to<br/>explain the behavior of return differentials and asset prices by suggesting the characteristics of the “marginal<br/>investor” in asset markets, or by pointing towards a direction in wh