样本问题

We give several sufficient conditions of NA and NSD properties of random variables, compare the closeness of distribution functions of order statistics between NA and independent rv's, investigate properties of NSD defined by supermodular functions, study the dependence structure of stationary Markov process and order statistics and their spacings from two samples, establish stochastic comparisons of order statistics of heterogeneous rv's in the hazard rate and the reversed hazard rate orders, and give the first analytic proof of the closure property of the up shifted likelihood ratio order under convolution.

给出判定一组随机变量NA性质的几个充分条件，并给出NA和独立随机变量的次序统计量之间"贴近性"的比较，系统地研究基于特殊函数类所定义的NSD负相依概念，给出一些判别NSD性质的结构定理和一些有用概率不等式。系统研究平稳马氏过程的相依结构。研究两样本次序统计量及其问题隔的相依结构。在失效率、反向失效率序和似然序意义下给出非齐次随机变量次序统计量的比较。给出上漂移似然序卷积封闭性的第一个解析证明。

In consideration of endogeneity problems, we use the trivariate probit model which estimate three equations- employment, liver disease and drinking . Finally we used WESML (weighted exogeneous sampling maximum likelihood) method to correct the bias caused by choosing specific sample.

其中，我们考量了肝病与饮酒习惯分别对於就业的内生性问题，并且在trivariate probit模型中使用工具变数来得到一致性的估计结果，最后，我们利用WESML (weighted exogeneous sampling maximum likelihood)的估计方式来修正因为选取某特定样本，所造成估计结果缺乏代表性的问题。

Moreover, we examine how the agency cost between managers and liquid shareholders is affected by nonliquid shares and stateowned shares, which we believed to exist only in Chinese Market. Through the same method of event study, we find a symmetric effect regardless of the proportion of nonliquid shares, and this shows that the proportion of nonliquid shares is not a factor of conflicts between managers and liquid shareholders. As for the sub-sample with stateowned shares, there is a larger positive price impact for firms with high proportion of stateowned shares, which tells us that the stateowned shares do raise the problem of "absence of the owner", and it is a part of agency cost in China.

我们进一步检验了国内特殊存在的非流通股以及国有股对于管理层与流通股东之间代理成本的影响，通过同样的事件研究方法对上述样本进行市场宣告反应检验，结果发现非流通股比例不同的公司其市场反应并无显著差异，说明非流通股的存在对管理层与流通股东之间的代理问题没有影响；而国有股比例不同的公司其市场宣告反应存在显著差异，这说明国有股产生的&所有者缺位&问题成为国内代理成本的一部分，并确实影响了现金股利的发放。

The main contributions of this thesis are as follows:(1)To address the problem of kernel parameterσselection and regulation of the decision boundary in SVDD algorithm,this thesis proposes a new kernel parameter optimization method based on the spheral distribution of samples in kernel space and regulation of the decision boundary method based on KPCA(Kernel Principal Component Analysis).

本论文主要以支持向量数据描述(Support Vector Data Description，SVDD)与随机森林(Random Forests，RF)模式识别工具为基础，对流程工业在线故障诊断的若干问题进行研究，其具体内容如下：(1)针对SVDD的核参数σ优化及其决策边界规整问题，提出了基于核样本球形分布的核参数优化方法与基于核主元分析(Kernel Principal ComponentAnalysis，KPCA)的SVDD决策边界规整方法。

The shortages of typical BP algorithm are that its convergence speed is slow and it possibly mires to local minimum. To solve the former shortage, BP algorithm was modified from several aspects, such as adjusting learning rate adaptively, adding momentum factor, preprocessing training data and modifying network structure, etc.. For the latter shortage, factor γ was introduced to avoid most of local minimum and accelerate the convergence speed.

为解决现有BP算法收敛速度慢的问题，对BP算法进行了改进，采用了动态改变步长、加入动量因子、对训练样本进行预处理以及调整网络结构等措施；针对经典BP算法可能收敛到局部极小点的问题，加入γ因子，可以避免大部分的局部极小值，从而使网络的收敛速度加快。

At last, the research test the logistic regression model with sample dada. The results are as the following:(1)Based on the time series dada, the financial early-warning model contains more crucial information reflecting the dada characters, so it can avoid the obvious logic problem in the other analysis such as single section dada and multi-section dada analysis. The test result indicate that the new model have a more accuracy than others.(2) the research adopt the global principal component analysis to settle the problem of multiple collinearity, so the model would be more normative and precise.(3)the early-warning model uses the backward method to regress, so it gets a terser model of only five variables, but it still has a precise forecasting ability.

通过本文研究说明：（1）建立以时序数据为基础的财务危机预警模型，提高了样本中重要财务趋势特征的信息含量，从而克服了国内外单截面及多重截面分析中普遍存在的明显逻辑问题，检验结果表明，利用时序数据建立的logistic财务危机判别模型较截面数据基础模型的判别准确率要高；（2）采用全局主成份分析方法解决了模型中因子间的多重共线性问题，这一处理使logistic回归方法应用的规范度和严谨度得到明显改善；（3）我们采用向后逐步回归的办法进行回归，使模型大大得到了简化，其中只包含五个自变量，但是它的整体预测准确性却仍然比较高，使模型更具有实用性。

The author ' s opinions related to the automatic sprinkling fire system design such as the nozzles, pipeline network, installation of alarm valve and the protection of the system are proposed on the basis of personal experiences in practice for years.

根据多年工作经验，就自动喷水灭火系统的喷头、管网、报警阀设置、系统的保护面等问题提出己见。并介绍CO2气体灭火系统，新型灭火剂FM-200和烟烙尽在应用中应注意的一些问题。关键词：自动喷水灭火系统 CO2气体灭火系统 FM-200 烟烙尽--样本内容

The pattern recognition problems with large deformations of samples is challenging in the field of pattern recognition at present.

目前在模式识别领域中，模式样本变化大的模式识别问题是仍未解决的问题之一。

This paper describes a new method for constructing a rough neural network. In this network, rough neurons lie in the hidden-layer, and they consist of three parts which generated by two hyperplanes that partition the universe.

本文根据支持向量机可以解决小样本学习问题的优势，再结合粗集理论对不确定性问题分析的特点，提出一种支持向量机的粗糙神经网络的构造方法。

The conclusion that dissolved gases in gas composition and content of transformer could embody the inside of the transformer"s situation about "the insulation fault effectively is concluded.In the research of power equipment fault diagnosis based dissolved gases in transformer oil, the fault type increase clearly with the fault development. Traditional way is inability to faults diagnose of electrical equipment such as transformer that the fault reason, phenomenon and principle are indeterminacy etc.

在以油中溶解气体分析电气设备故障的模式识别研究中，当故障进一步发展，待判别的类型将显著增加，由于传统的诊断方法缺乏对样本数据的科学分析，不能够真实体现特征量数据的规律，将不可避免的面临准确率下降的问题，且不可能提供类间相互关系或演变过程的信息，因此用新的诊断方法来解决此类问题十分必要；本文在对目前变压器内部绝缘故障常用诊断方法总结分析的基础上，利用模糊数学的思想，提出以油中溶解气体为特征量，采用模糊聚类的方法对变压器内部绝缘故障进行诊断。

I can not make it blossom and suits me

我不能让树为我开花

When temperatures are above approximately 80 °C discolouration of the raceways or rolling elements is a frequent feature.

当温度高于 80 °C 左右时，滚道或滚动元件褪色是很常见的特征。