Bayesian median autoregression for robust time series forecasting

We develop a Bayesian median autoregressive (BayesMAR) model for time series forecasting. The proposed method utilizes time-varying quantile regression at the median, favorably inheriting the robustness of median regression in contrast to the widely used mean-based methods. Motivated by a working Laplace likelihood approach in Bayesian quantile regression, BayesMAR adopts a parametric model bearing the same structure as autoregressive models by altering the Gaussian error to Laplace, leading to a simple, robust, and interpretable modeling strategy for time series forecasting. We estimate model parameters by Markov chain Monte Carlo. Bayesian model averaging is used to account for model uncertainty, including the uncertainty in the autoregressive order, in addition to a Bayesian model selection approach. The proposed methods are illustrated using simulations and real data applications. An application to U.S. macroeconomic data forecasting shows that BayesMAR leads to favorable and often superior predictive performance compared to the selected mean-based alternatives under various loss functions that encompass both point and probabilistic forecasts. The proposed methods are generic and can be used to complement a rich class of methods that build on autoregressive models.     

Estimating survivor function using optimally selected order statistics

This paper deals with the estimation of survivor function using optimally selected order statistics when the sample sizen is large. We use the estimates (μ*,σ*) based on the optimum set of order statistics for largen and fixedk (≤n) such that the estimate has optimum variance property. The asymptotic relative efficiency of such an estimator is compared with the one based on the complete sample. The general theory of the problem and specific details with respect to a two-parameter Normal, Logistic, Exponential and Pareto distributions is considered as an example.     

Dependence structure and symmetry of Huang-Kotz FGM distributions and their extensions

An extension of FGM class of bivariate distributions with given marginals is presented. For Huang-Kotz FGM distributions some theorems characterizing symmetry and conditions for independence are obtained. The new family of distributions allows us to achieve correlation between the components greater than 0.5.     

Extremes and Robustness: A Contradiction?

Stochastic models play an important role in the analysis of data in many different fields, including finance and insurance. Many models are estimated by procedures that lose their good statistical properties when the underlying model slightly deviates from the assumed one. Robust statistical methods can improve the data analysis process of the skilled analyst and provide him with useful additional information. For this anniversary issue, we discuss some aspects related to robust estimation in the context of extreme value theory (EVT). Using real data and simulations, we show how robust methods can improve the quality of EVT data analysis by providing information on influential observations, deviating substructures and possible mis-specification of a model while guaranteeing good statistical properties over a whole set of underlying distributions around the assumed one.     

地铁盾构隧道洞门环中心测量方法及应用研究

地铁盾构隧道洞门环是盾构机始发和接收的主要载体,需要精确地测定洞门环中心的三维坐标,为相关工作提供数据支撑。在工程中普遍采用的方法是在洞门环水平放置安装有管水准气泡的水准尺,采用全站仪采集其中心坐标,并利用其和洞门环的空间几何关系间接求出洞门环中心空间三维坐标。论文以青岛地铁4号线九静区间盾构始发井洞门环中心测量为例,介绍了采用全站仪实测三维坐标进行洞门环中心拟合的方法,并对两种方法的优缺点进行了比较分析,可为类似工程的测量提供有益的借鉴参考。     

房室模型(Compartment Model)在权益变动分析中的应用

本文将负债和业主权益看成是房室模型中的两间房室,将负债与业主权益的变动问题转化为房室模型的求解问题,并以会计理论与经济学理论为依据,设定假设,使两房室模型中的6个参数求解成为可能,从而发现了权益的时间分布规律。最后本文利用一家上市公司会计报表的历史数据,验证了对权益变动建立房室模型进行求解及预测的可行性。     

协方差结构模型不收敛原因及解决方法

协方差结构模型实际上是一般线性模型的扩展,它主要利用一定的统计手段对复杂的理论模式加以处理,并根据各种拟合指数对模型作出检验与评价,从而达到证实或证伪事先假设的理论模型的目的。笔者在利用协方差结构模型对北京市居民住房消费行为和意愿进行量化研究时发现协方差结构模型存在不收敛问题,文章提出,导致模型不收敛的原因,一是缺失数据处理方法不当,可采用期望最大化算法(EM算法)和马尔科夫链蒙特卡罗法(MCMC算法)处理数据缺失;二是变量间存在多重共线性,可去掉设置不合理的潜变量以避免共线性问题;三是模型过于复杂,收敛条件苛刻,可调整模型使之简单化,并重新设定收敛条件,促使模型收敛。     

A Discrete Time Equivalent Martingale Measure

An equivalent martingale measure selection strategy for discrete time, continuous state, asset price evolution models is proposed. The minimal martingale law is shown to generally fail to produce a probability law in this context. The proposed strategy, termed the extended Girsanov principle, performs a multiplicative decomposition of asset price movements into a predictable and martingale component with the measure change identifying the discounted asset price process to the martingale component. However, unlike the minimal martingale law, the resulting martingale law of the extended Girsanov principle leads to weak form efficient price processes. It is shown that the proposed measure change is relevant for economies in which investors adopt hedging strategies that minimize the variance of a risk adjusted discounted cost of hedging that uses risk adjusted asset prices in calculating hedging returns. Risk adjusted prices deflate asset prices by the asset's excess return. The explicit form of the change of measure density leads to tractable econometric strategies for testing the validity of the extended Girsanov principle. A number of interesting applications of the extended Girsanov principle are also developed.     

Generalized Hyperbolic Diffusion Processes with Applications in Finance

A special class of diffusion processes, the generalized hyperbolic diffusion processes, is introduced. As a byproduct we present a technique for the construction of one-dimensional ergodic diffusion processes with a predetermined stationary density. We specifically study the application of this new type of diffusion process to financial data, especially U.S. stock prices. It is seen that in addition to confirming stylized features of the financial market, a key explanation concerning "thick-"tailed log returns is provided.     

Definition of full-time and part-time employment,and distributional assumptions: the implications for the estimated full-time and part-time wage equations

This paper demonstrates that rejecting the standard definition of full-time and part-time workers, the estimated number of hours that an individual is likely to work as a full-time worker is a function of the type of distribution one assumes about the error term in the wage equation. Adopting a switching regression model with unknown sample selection, we have found that the normality assumption generates higher hours for full-timers in comparison with the non-normal distributions. We also noted that regardless of the distribution assumed, the hours differ from one industry to another. The implication is that the standard definition of full-time and part-time worker may not be appropriate for all firms irrespective of the distribution assumed. The paper also shows the sensitivity of parameter estimates to the distributional assumptions about the error term in the wage equation. The results indicate that the normal distribution wage equation estimates are relatively larger than the Weibull and exponential distributions. This finding is particularly important because such differences in estimated coefficients may have a direct wage influence on the wage gap between full-time and part-time workers across distributions.