基于贝叶斯法则的空间自相关误差自相关模型变量选择研究

本文将研究贝叶斯法则视角下的空间自相关误差自相关模型(Spatial Autoregressive Model with Autoregressive Disturbances,SARAR模型)变量选择问题。通过将基于BIC准则的子集选择法推广到空间模型,实现SARAR模型的变量选择,并证明在一定条件下,对于SARAR模型的变量选择BIC准则具有良好的渐近性质。同时本文还将利用Monte Carlo模拟验证BIC准则能够很好的实现SARAR模型的变量选择。最后以股票收益率为例,在验证股票收益率具有空间效应的前提下,利用BIC准则对影响股票收益率的众多财务指标进行变量选择。     

基于Spatial AIC准则的空间自回归模型变量选择研究

变量选择直接决定着空间计量经济模型的有效程度与实证研究结果。为有效解决空间自回归模型(即SAR模型)的变量选择问题,本文利用Kullback-Laible信息量最大化,把AIC准则运用到SAR模型构建,推导出Spatial AIC统计量,提出Spatial AIC准则。然后利用统计理论证明Spatial AIC准则选择SAR模型变量的渐近最优性;利用蒙特卡洛模拟方法,比较Spatial AIC准则、经典AIC准则和Lasso方法用于SAR模型变量选择的有限大样本性质;利用空间相关的沪深300成分股股票收益率数据,采用Spatial AIC准则和Lasso方法,分别构建股票收益率财务因素的空间自相关模型,实证比较其相对有效性。三种结果均表明Spatial AIC准则能够更好地解决SAR模型变量选择问题。     

半参数GAM模型下我国教育收益率的分析

本文利用广义可加模型拓展了明瑟人力资本方程,应用于(《中国健康营养调查(CHNS)》中我国城乡居民8年的数据研究,发现教育收益率具有以下变化规律:一是2004年我国的个人教育收益率达到最大值。受我国高等教育大众化的影响,2006、2009年个人教育收益率持续下降;二是获得最高个人教育收益率的教育阶段沿着完成小学、初中、高中、大学专科,即朝更高教育水平的方向缓慢移动;三是工作经验年数与受教育年限不具有"平行性",个人教育收益率的变化受工作经验年数的显著影响。     

高维线性回归模型稳健变量选择方法综述

随着大数据时代的到来,在经济学、金融学和生物医学等众多研究领域中频繁收集到高维数据.高维数据的特征之一是变量维数p随着样本量n的增加而变大且通常会超过样本量,同时,异常值也容易出现在高维数据中.因此,如何克服异常值给高维统计推断带来的影响,从而得到更精确的模型,是目前统计学研究的热点问题之一.本文是对高维线性模型下的稳健变量选择方法进行综述.具体地,首先介绍评估稳健性的三个指标:影响函数、崩溃点和最大偏差.其次着重介绍了稳健变量选择方法,包括响应变量含有异常值,响应变量和协变量都含有异常值,高崩溃点且高效的变量选择方法.紧接着介绍相关算法,通过模拟和实例比较不同变量选择方法.最后,简要探讨了高维稳健有效变量选择方法存在的问题及未来的可能发展方向.     

GARCH模型估计方法选择及对上证指数的应用

本文介绍了对ARCH/GARCH模型的两种估计方法:准极大似然估计和极小绝对偏差估计,并提出了一种基于自助法(Bootstrap)对估计方法的选择。在厚尾程度不同的情况下进行了模拟分析,表明对于一个具体的数据,该选择法能够自动选择较优的估计方法。并用该方法对上海证券交易所A股和B股的股价指数进行了分析,印证了上海股市B股收益率的尾部厚于A股收益率尾部。     

不同风险度量方法的投资组合比较研究

为了验证投资组合理论在中国证券市场的有效性,在不允许卖空情况,针对不同风险度量方法,文章运用旋转算法或结合序列二次规划法分别求解均值-方差、均值-下半方差投资组合模型、均值-半绝对偏差、均值-平均绝对偏差和均值-VaR.文章选取三年沪市六只业绩比较好的股票,依据前两年的数据作为样本数据,分别求出五个模型在不同期望收益率下的最优投资策略,将得出的最优投资策略应用到最后一年,进行模拟投资,从而计算出各模型的总收益率.以等比例投资为标准,比较五个模型的绩效.最后,证明了两个模型对于中国证券市场是适用.     

不允许买空时的均值-下方风险投资组合选择——基于非参数估计方法

本文分别利用下半方差和下半偏差度量下方风险(Downside Risk),采用非参数估计方法研究了不允许买空时的均值-下方风险投资组合选择问题。首先,我们利用组合收益率密度函数的非参数核(kernel)估计得到了下方风险的计算公式,并把它们嵌入到均值-下方风险投资决策模型中。然后得到了模型最优解存在的充要条件,并给出了求解最优投资策略的Zoutendijk可行方向算法。最后,基于中国股票市场真实数据给出了一个数值算例,并比较了在两种不同下方风险度量下模型的差别。     

稀疏VAR在股票收益率研究的应用

向量自回归模型(VAR)广泛应用在对时间相依的多元时间序列建模中,但在高维数据建模中,自回归的系数膨胀可能导致噪音估计、不稳定的预测、解释上的困难等问题。在实际应用中,序列的真实模型往往具有稀疏性,因此运用稀疏VAR模型对高维时间序列进行建模,不仅可以解决高维数据带来的上述困难,也有利于寻找高维数据内在的真实模型。本文以10家公司的股票收益率为研究对象,采用3种不同的稀疏估计方法,不但分析了股票收益率之间的动态关系,而且通过实证分析展示了稀疏估计的优势。     

纵向数据下线性EV模型的变量选择

本文考虑了纵向数据线性EV模型的变量选择.基于二次推断函数方法和压缩方法的思想提出了一种新的偏差校正的变量选择方法.在选择适当的调整参数下,我们证明了所得到的估计量的相合性和渐近正态性.最后通过模拟研究验证了所提出的变量选择方法的有限样本性质.     

半参数平滑转换分位数自回归模型及应用

已有针对平滑转换自回归模型(STAR)的研究多是将转换函数设定为Logistic函数或指数函数形式,并在均值回归框架下获得模型的估计、检验及预测结果.文章基于重心权有理插值和分位数回归方法,构建一类新的半参数平滑转换分位数自回归模型,其主要特点表现在:第一,基于重心权有理插值方法构造的平滑转换函数,形式更加灵活自由,有效减少了模型误设的风险.第二,在分位数回归框架下,利用遗传算法获得新模型在不同分位点处的平滑转换自回归系数估计,比单纯的均值回归得到的信息更为丰富.数值模拟结果显示,新模型的平滑转换自回归系数估计在无偏性、有效性和一致性方面均具有较好表现.最后,将新模型应用于上证综指日收益率的动态趋势及预测研究,细致揭示了收益率序列在不同阶段、不同分位点处的非线性和异质性变化特征.     

二阶随机占优约束下考虑偏度的投资组合模型

在DentchevaRuszczynski(2006)模型的基础上,考虑偏度对构建投资组合的影响,建立了二阶随机占优约束下最大化组合收益率偏度的投资组合优化模型,并应用分段线性近似方法将模型转化为一个非线性混合整数规划问题.利用中国股票市场的历史数据对所建模型进行了实证分析,结果表明,所建新模型比均值-方差-偏度模型和市场指数具有更稳健的表现.     

A capital allocation based on a solvency exchange option

In this paper we propose a new capital allocation method based on an idea of [Sherris, M., 2006. Solvency, capital allocation and fair rate of return in insurance. J. Risk Insurance 73 (1), 71-96]. The proposed method explicitly accommodates the notion of limited liability of the shareholders. We show how the allocated capital can be decomposed, so that each stakeholder can have a clearer understanding of their contribution. We also challenge the no undercut principle, one of the widely accepted allocation axioms, and assert that this axiom is merely a property that certain allocation methods may or may not meet.     

An Extended FE Formulation for Elasto-Plastic Materials

We present an alternative method for the numerical simulation of elasto-plastic material behaviour. For this extended Finite Element (FE) formulation the history variables, which provide the information of plastic deformations from the previous timesteps, are represented as FE functions. This results in additional degrees of freedom (DOF), and the radial return of the standard formulation is replaced by a fully coupled Newton method for the extended system. Numerical studies, using viscoplastic regularization within a geometrically linear approach prove comparative results and an advantage in calculation time for the extended FE formulation. (© 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)     

基于DEA模型的中国家电上市公司效率实证分析

基于规模报酬可变的DEA模型和Malmquist指数方法,利用中国22家家电上市公司2006-2011年间的截面数据和面板数据,从静态视角和动态视角双维度考察了这一时期中国家电上市公司效率状况.研究发现,2006-2011年期间中国家电上市公司技术效率状况不佳,平均值为0.87,技术效率经历了下降后缓慢上升的趋势;中国家电上市公司动态效率较低,平均增长率为2.3%,且动态效率具有明显的波动性,缺乏持续的增长趋势,技术进步是中国家电上市公司动态效率增长主要动力,技术效率对其增长贡献较小.     

A study of leveraged ETF based on geometric return

Leveraged (inverse) exchange-traded funds (LETFs) seek to deliver multiples (opposite) of the performance of the index or benchmark they track. LETFs typically are designed to achieve their stated performance objectives on a daily basis. Many real-life and hypothetical examples have been given to show that the performance of these ETFs over a period longer than one day can differ from their stated daily performance objectives. Formulae have been found using both continuous method and discrete method. A discrete method was used to find a formula linking the return of a leveraged fund with the corresponding multiple of the return of the unleveraged fund and its realized variance but the method needs to use some assumptions and statistical properties to create the volatility term. A CME report finds a very simple way to include volatility in their formula but fails to link to the return of the corresponding unleveraged product. In this paper, we find a natural way to link a leveraged fund with its corresponding unleveraged product and its realized variance in a discrete manner. Our derivation process is similar to that in the CME report, so we do not need to use assumptions and statistical properties to create the volatility term. Unlike the CME method, we use geometric return as opposed to arithmetic return. So, we are able to connect with the return of the corresponding unleveraged product.     

Kinks in the STEM Pipeline: Tracking STEM Graduation Rates Using Science and Mathematics Performance

In an effort to maintain the global competitiveness of the United States, ensuring a strong Science, Technology, Engineering and Mathematics (STEM) workforce is essential. The purpose of this study was to identify high school courses that serve as predictors of success in college level gatekeeper courses, which in turn led to the successful completion of STEM degrees. Using a purposive sample of 893 students who had declared a STEM major between the fall of 2006 and the spring of 2008, data were collected on students' high school grades, college grades, national test scores, grade point average, gender, and ethnicity. Using analysis of variance, correlations, multiple discriminant function analysis, and multiple regression models we found that high school calculus, physics, and chemistry (respectively) were predictors of success in STEM gatekeeper college courses. Then using those courses, we constructed a predictive model of STEM degree completion. The implications of this study highlight and reinforce the importance of providing rigorous mathematics and science courses at the high school level, as well as provide some evidence of a potential mediated model of the relationship between high school performance, college performance, and graduating with a STEM degree.     

Estimating the Positive and Negative Jumps of Asset Returns Via Kalman Filtering. The Case of Nasdaq Index

The daily asset (log) return is considered to consist of two parts, the positive and negative jump. These jumps are determined by the arrival of positive and negative news in the market, are not observable and their differences define the asset returns. In order to estimate the jumps, the basic discrete time-homogeneous linear Kalman filter is applied, in which all the noises are assumed to be normally distributed. Then, under the assumption that the estimated jumps have to be non-negative, the method of their pdfs’ truncation, according to the non-negativity constraints, is used, and it is accompanied by appropriate scaling. The fitting of the model is justified by examining the fitting of the estimated returns to the empirical ones. For that purpose, the daily Nasdaq returns during the three year period 2006-2008 are used.     

基于非参数分布的沪深股市VaR测度

沪深大盘指数的收益率分布函数并不服从通常人们所认为的正态分布.因此,采用一种新的方法—非参数核密度估计,对沪深股指收益率分布进行拟合.该方法不仅很好地刻画了收益率分布的尖峰和肥尾特征,而且由此建立的VaR模型比一般的基于参数分布的VaR模型更能捕捉市场的风险特征,结论也更加准确.     

沪深股市收益分布尾部特征研究

本文运用跳跃扩散模型和极值理论方法对沪深股指收益分布特征进行了研究。跳跃扩散模型定量化地给出沪深股市股票收益分布产生的原因,沪深股市收益分布为具有胖尾的非正态分布,股票收益变动主要是由离散信息作用引起,一般帕累托分布较好地拟合股票收益左尾分布。     

用矩匹配方法计算非线性风险度

矩匹配方法是用来求解非线性风险度(Value at Risk，简称：VaR)的一种普遍性方法，它是先假定样本经验分布服从已知分布族，然后运用矩匹配估计方法估计相应的参数，得到资产回报样本的密度函数，再计算风险度VaR；本采用的Johnson分布族是矩匹配方法的直接应用，并且计算出来的结果与局部Monte．Carlo结果进行了比较。并通过实证分析认为这种方法是一种良好的计算非线性VaR方法。