偏正态数据下联合位置与尺度模型的统计诊断

本文研究偏正态数据下联合位置与尺度模型,考虑基于数据删除模型的参数估计和统计诊断,比较删除模型与未删除模型相应统计量之间的差异.首次提出基于联合位置与尺度模型的诊断统计量和局部影响分析.通过模拟研究和实例分析,给出不同的诊断统计量来判别异常点或强影响点,研究结果表明本文提出的理论和方法是有用和有效的.     

基于Bayes方法的诊断模型分析

文中证明了在Bayes框架下,当漂移参数7服从有信息先验时,在相当广泛的统计模型中,数据删除模型(CDM)和均值漂移模型(MSOM)的参数估计不相等,几个数值例子验证了相应的结论.     

基于模糊Cook距离的模糊线性回归模型的统计诊断

针对确定输入、模糊输出的模糊线性回归分析模型,采用最小二乘法,讨论了模糊线性回归模型的数据删除模型的参数估计,将建立在确定性数据基础上的线性回归模型统计诊断量Cook距离推广到模糊线性回归分析模型中,构造了统计诊断量—模糊Cook距离,通过数值模拟和对实际例子的研究,识别出其中的强影响点,得出与其它方法相同的结论,表明本文构造的统计诊断量是有效的,且应用比其它方法更方便.     

m维AR(p)模型的统计诊断

统计诊断就是探查对统计推断（如估计或预测等）有较大影响的数据从而对全过程数据进行诊断．本文应用基于数据删除模型得到二维AR（1）模型的参数估计诊断公式,给出了Cook统计量的计算公式,进而推广到m维AR（p）模型的情形．     

ZI数据的统计分析综述

ZI (zero-inflated)数据就是含零过多的数据.从上世纪90年代以来, ZI数据在各个研究领域受到越来越广泛的重视,现在仍然是数据分析的热点问题之一.本文首先通过2个实例说明ZI数据的实际意义,然后介绍ZI数据分析的研究概况和最新进展.另外文章还系统介绍了各种ZI数据模型、ZI纵向数据模型及其参数估计方法,同时也介绍了ZI数据的统计诊断等问题, 其中包括作者近年来的一些工作.最后, 本文列出了若干有待进一步研究的问题.     

具有椭球误差的线性一致相关模型的相关性检验

本文讨论具有椭球误差的线性一致相关模型的相关性检验问题.基于Fisher-score方法给出模型参数估计的迭代公式,然后分别对一致相关系数进行了存在性和齐性检验,得到了相应检验的score统计量,同时给出了功效模拟.最后利用实际数据说明了模型以及检验统计量的价值.     

基于Pena距离的双重广义线性模型的统计诊断

本文主要研究双重广义线性模型,考虑基于数据删除模型的参数估计和统计诊断,比较删除模型与未删除模型相应的诊断统计量之间的变化.首次提出基于双重广义线性模型下的Pena距离.通过一些模拟研究以及实例分析,比较不同诊断统计量判别异常点或强影响点的差异,研究结果表明本文提出的理论和方法是行之有效的.     

基于杠杆值大数据集抽样的异常点诊断

本文主要研究大数据集下利用杠杆值抽样后的异常点诊断问题。首先讨论了数据删除模型中参数估计的统计性质,构造了四种异常点诊断统计量;其次,根据均值漂移模型的漂移参数的假设检验问题,构造了三种检验统计量;最后,通过模拟和实证数据分析结果得出本文的结论—异常点诊断对于基于杠杆值的大数据集抽样估计起到重要的影响作用。     

成分数据的线性模型

本文建立了成分数据的线性模型,对参数给出了估计,证明了参数估计的多种性质.     

考虑高频数据影响的BEKK模型的估计及其应用

多维资产的协方差阵在投资组合中扮演着重要角色,如何估计和预测资产的协方差阵是统计领域的一大热点问题.将基于高频数据的已实现协方差阵(RCOV)和双频已实现协方差阵(TSCOV)应用到BEKK模型的估计过程中,提出了考虑高频数据影响的BEKK-RCOV和BEKK-TSCOV模型,这两类模型将高频数据引入到协方差阵估计过程中的同时,还可以对协方差阵直接进行预测,避免了预测模型的选择困难问题,并且提高了协方差阵的估计效率.通过实证研究发现:BEKK-RCOV和BEKK-TSCOV模型估计和预测效果明显优于BEKK模型,将其应用在投资组合时,使投资者获得了更高的收益.     

Fitting continuous-time and discrete-time models using discrete-time data and their applications

An exponential function scheme, which is an extension of the time-domain prony method, and a mixed-matching method are developed for fitting the coefficients of both continuous-time and discrete-time transfer functions, using the discrete-time data of either continuous-time or discrete-time systems. When the discrete-time data are obtained from a continuous-time (discrete-time) system and the discrete-time (continuous-time) models are desirable, the proposed method can be applied to perform the model conversions. If the discrete-time data are obtained from a high-degree system, the proposed method can be applied to determine the reduced-degree models.     

Integration of generative and evaluative models for production scheduling of lube oil plants in a petroleum refinery

A study was made on the existing practices of production planning, scheduling and prevailing constraints in the six plants of a lube oil section in a petroleum refinery. Based on the data collected from these plants, some generative and evaluative models were developed. The generative models developed were flow network optimisation (FNO) model and binary integer linear programming (BILP) model. The evaluative model developed was simulation. The optimal results obtained from the generative model were fed to the evaluative model to derive the measure of performance. This integration of generative and evaluative models offers an opportunity for better understanding of the subsystem and appropriate decision making.     

多变量整体模式累加多层统计模型的建立及其在组织绩效预测上的应用研究

为了解决多层的少样本或无规则数据的建模问题,在一般多层统计模型的基础上提出了多变量整体模式的累加多层统计模型。此模型把累加方法的优点(将无规则数据转化成有规则数据)与多层统计模型结合起来,拓展了多层统计模型的适用范围。从其在香蕉组织绩效的分析以及在仅有两个调查数据香蕉组织形式绩效的预测中,可以看出此模型有较强的实用性。     

Spline estimation of functional coefficient regression models for time series with correlated errors

In this work we focus on functional coefficient regression (FCR) models. Here we study the estimation of FCR models by splines, with autoregressive errors and show the rates of convergence of the proposed estimator. The importance of taking into account the correlation is assessed via simulation studies and multi-step ahead forecasts for a real data set.     

A semiparametric panel approach to mortality modeling

During the past twenty years, there has been a rapid growth in life expectancy and an increased attention on funding for old age. Attempts to forecast improving life expectancy have been boosted by the development of stochastic mortality modeling, for example the Cairns–Blake–Dowd (CBD) 2006 model. The most common optimization method for these models is maximum likelihood estimation (MLE) which relies on the assumption that the number of deaths follows a Poisson distribution. However, several recent studies have found that the true underlying distribution of death data is overdispersed in nature (see Cairns et al. 2009 and Dowd et al. 2010). Semiparametric models have been applied to many areas in economics but there are very few applications of such models in mortality modeling. In this paper we propose a local linear panel fitting methodology to the CBD model which would free the Poisson assumption on number of deaths. The parameters in the CBD model will be considered as smooth functions of time instead of being treated as a bivariate random walk with drift process in the current literature. Using the mortality data of several developed countries, we find that the proposed estimation methods provide comparable fitting results with the MLE method but without the need of additional assumptions on number of deaths. Further, the 5-year-ahead forecasting results show that our method significantly improves the accuracy of the forecast.     

Structural-EM for learning PDG models from incomplete data

Probabilistic Decision Graphs (PDGs) are a class of graphical models that can naturally encode some context specific independencies that cannot always be efficiently captured by other popular models, such as Bayesian Networks. Furthermore, inference can be carried out efficiently over a PDG, in time linear in the size of the model. The problem of learning PDGs from data has been studied in the literature, but only for the case of complete data. We propose an algorithm for learning PDGs in the presence of missing data. The proposed method is based on the Expectation-Maximisation principle for estimating the structure of the model as well as the parameters. We test our proposal on both artificially generated data with different rates of missing cells and real incomplete data. We also compare the PDG models learnt by our approach to the commonly used Bayesian Network (BN) model. The results indicate that the PDG model is less sensitive to the rate of missing data than BN model. Also, though the BN models usually attain higher likelihood, the PDGs are close to them also in size, which makes the learnt PDGs preferable for probabilistic inference purposes.     

Sequential and non-sequential acceptance sampling plans for autocorrelated processes using ARMA(p,q) models

This paper presents variable acceptance sampling plans based on the assumption that consecutive observations on a quality characteristic(X) are autocorrelated and are governed by a stationary autoregressive moving average (ARMA) process. The sampling plans are obtained under the assumption that an adequate ARMA model can be identified based on historical data from the process. Two types of acceptance sampling plans are presented: (1) Non-sequential acceptance sampling: In this case historical data is available based on which an ARMA model is identified. Parameter estimates are used to determine the action limit (k) and the sample size(n). A decision regarding acceptance of a process is made after a complete sample of size n is selected. (2) Sequential acceptance sampling: Here too historical data is available based on which an ARMA model is identified. A decision regarding whether or not to accept a process is made after each individual sample observation becomes available. The concept of Sequential Probability Ratio Test (SPRT) is used to derive the sampling plans. Simulation studies are used to assess the effect of uncertainties in parameter estimates and the effect of model misidentification (based on historical data) on sample size for the sampling plans. Macros for computing the required sample size using both methods based on several ARMA models can be found on the author’s web page .     

Air path identification of diesel engines by LPV techniques for gain scheduled control

Modelling is a key element to improve the performance of engine control systems, but many factors like non-linearity and complexity complicate the derivation of sufficiently precise physical models. This motivates an increasing interest in data based models. Linear models can successfully represent the engine operation in some reduced regions, but tend to fail when large operating regions must be considered. This motivates the interest in deriving and using gain scheduling models or their natural extension, the linear parameter varying (LPV) models. In this article we propose to model the air path of diesel engines using input–output LPV models with a physically motivated structure and parameters estimated from data. These models are shown to combine good precision with simplicity and allow the systematic design of optimal and robust control systems, and can be determined in a very short time if sufficient data are available.     

多元测量误差模型的稳健GM-估计量

由于EV（Errores-in-Variables)模型（也称测量误差模型）的最大似然估计由正交回归给出，而正交回归对污染数据是敏感的，所以，需要采用稳健的统计方法来估计模型参数。本文在多元EV模型中引入稳健GM-估计量，把一元正态EV模型的若干结果推广到多元情形，所得的稳健性结果不仅更具一般性，而且还修正了文献中对一元情形给出的一个错误结果。     

Evaluating the goodness of fit of stochastic mortality models

This study sets out a framework to evaluate the goodness of fit of stochastic mortality models and applies it to six different models estimated using English & Welsh male mortality data over ages 64-89 and years 1961-2007. The methodology exploits the structure of each model to obtain various residual series that are predicted to be iid standard normal under the null hypothesis of model adequacy. Goodness of fit can then be assessed using conventional tests of the predictions of iid standard normality. The models considered are: Lee and Carter’s (1992) one-factor model, a version of Renshaw and Haberman’s (2006) extension of the Lee-Carter model to allow for a cohort-effect, the age-period-cohort model, which is a simplified version of the Renshaw-Haberman model, the 2006 Cairns-Blake-Dowd two-factor model and two generalized versions of the latter that allow for a cohort-effect. For the data set considered, there are some notable differences amongst the different models, but none of the models performs well in all tests and no model clearly dominates the others.