参数估计

The common method, that all strong-correlation terms of the model are eliminated, can bring the loss in the engineering application, so the new method is proposed that the identified model reserves some correlation. The augmented matrix A is constructed by the outputΔW and the matrix S. The"determinating order based on ratio of determinant"is brought out to screen the strong-correlation terms in the structure identification. The latent root estimation is improved in screening the eigenvalues and eigenvectors. Thus the estimation precision is improved greatly.The consistence check of guidance instrument error coefficients of flight test and ground test is the purpose of flight experiment. The causes of inconsistency of the two models are analyzed. The hypothesis test of linear regression model based on F statistics is proposed to check the consistence.Finally, the instability of error coefficients is probably caused by the change of the flight environments, therefore, the relation between the error coefficients and flight environment is analyzed. The approach is presented to identify SINS guidance instrument error models and compensate the error in the segmented sections corresponding to the change of vertical acceleration of aircraft.

在结构辨识中，常用的方法由于将模型中的强相关项全部剔除而给工程应用带来损失，因此，本文提出了新的有益思想，即在保留一定相关性的基础上进行辨识：将输出向量ΔW与环境函数矩阵S构成增广矩阵A，然后采用"比定阶行列式"来剔除相关向量的方法，这样既可以尽可能多地保留了对落点影响大的强相关参数，又可以对落点影响小的强相关参数给予剔除；在参数估计中，改进了特征根估计中特征根和特征向量的筛选方法，提出"近零"准则，从而大大提高了参数估计的精度；再者，鉴于天地模型"一致性"检验是飞行试验和SINS制导工具误差系数分离的主要目的，因此，本文又深入分析了造成天地模型不一致的原因，提出了采用基于F统计的线性回归模型假设检验方法来进行捷联制导工具误差模型的天地"一致性"检验；最后，鉴于飞行环境剧烈变化可能会对惯性仪表误差系数稳定性带来一定的影响，因此本文深入地分析了SINS制导工具误差系数与外界环境的关系，提出了基于过载变化大小的分段辨识和分段实时补偿的算法。

In this paper, we study characterizations of admissible in the general linear model Y, Xβ,ε|ε～(0,σ~2∑. We demonstrate that an admissible linear estimator is as the conditional generalized ridge-type estimation in the no constraint, equality constraint, inequality constraint general linear model. We study the superiority of this conditional generalized ridge-type estimation, and prove that it is superior to the restricted best linear unbiased estimator in terms of mean squares. We also give the choice of the matrix K.

本文主要研究了一般线性模型Y，Xβ，ε|ε～(0，σ~2∑中参数估计的可容许性特征，得到了一般线性模型在无约束，有等式约束及有不等式约束下，可容许线性估计均具有条件广义岭估计的形式的结论，并且讨论了这一条件广义岭估计的优良性，证明了其在均方误差和均方误差矩阵意义下都优于约束最小二乘估计，给出了参数矩阵K的选取方法。

This article estimates the extreme loss distribution with extreme value theory, estimates the shape parameter with HKKP estimator, which is unbiased to small sample data, selects the right threshold by selecting smallest mean square error of the estimated cumulative distribution function and empirical cumulative distribution function and based on the extreme value theory method, this article estimates the extreme loss distribution and the p-percentile operational loss under certain believe range for Chinese commercial banks, then calculates the operational risk supervise capital.

本文利用极值理论对中国商业银行操作损失极端值分布进行估计，针对尾参数估计的来用传统Hill佑计对小样本数据容易产生偏倚的情况，提出了采用Hill佑计的改进―小样本无偏估计的HKKP佑计来估计操作损失的尾参数，针对由于阈值确定不准确导致结果偏差大的情况，采用最小化估计的累计概率分布与经验累计概率分布平均平方误差的方法确定较精确的阈值，估计出给定置信水平下操作风险损失的分位数，从而使得中国商业银行操作风险监管资本的测定成为可能。

The nonlinearity and discrete gradient inherited in CAViaR model is a conundrum for parameter estimation. We take the asymmetric Laplace distribution with scale parameter as the error process; indicate the variance has a minimum positive value when the scale parameter is a constant, conflicting with the distribution of real financial data. Further we estimate the parameters of indirect TARCH-CAViaR model base on Bayesian framework and Markov chain Monte Carlo method. The optimal scale parameter can also be obtained by Markov chain Monte Carlo method.

CAViaR一般模型中递归分位回归方程的非线性和非连续可微性是参数估计的一个难题，基于含有尺度参数的不对称拉普拉斯分布作为误差过程，指出将尺度参数固定为常数会导致不对称拉普拉斯分布随机变量的方差存在最小正值的限制，与实际金融数据分布不符；进而提出采用贝叶斯分析和马尔科夫链蒙特卡罗模拟方法，估计间接TARCH-CAViaR模型的参数，并可获得尺度参数的合理估计。

The thesis analyzes the methods for parameter estimation of distributed sources in detail, and low complexity estimators, comes up with RELAX in parameter estimation and robust method of high resolution parameter estimators for distributed sources, and gives theory analysis and simulation results. The main work is:1 This text gives a study of the signal models of two kinds of distributed sources: coherent distributed sources and incoherent distributed sources, analyzes eigenvalue distribution of correlation matrix.2 Systematically introduced parameter estimation of distributed sources on the base of models, including the maximum likelihood estimate, least squares estimator, DSPE, DISPARE, etc.

论文详细分析了已有分布式目标参数估计方法及低复杂度的参数估计方法，提出了分布式目标参数估计的RELAX算法和具有稳健性的分布式目标高分辨估计方法，并给出了相应的理论分析和仿真实验结果。

Experimental results show that the performance of local parameter estimation is superior to that of global parameter estimation when there are no attacks, and the performance of local parameter estimation is worse than that of global parameter estimation when there are attacks.

通过实验得到以下结论：在无攻击条件下，局部参数估计的性能优于整体参数估计的性能；在有攻击条件下，整体估计的性能要好于局部参数估计的性能；压缩攻击会导致水印误检概率的增加，不同的攻击对模型造成的影响不同。

By using the least square method and the weighted functionmethod of nonparametric estimation together, this paper defines the wavelet estimators of β and g.

本文采用了半参数估计常用的办法之一——最小二乘法结合非参数权函数估计法定义了参数分量和非参数分量的估计，这里的权函数不是通常所采用的核权函数或近邻权函数，而是由小波方法所得到的权函数，我们称它为小波权函数。

In chapterⅡ, considering GLMMs data which the number of high units is different and the number of low units is small and unbalanced, we Conduct a series of simulations, the simulations show that regardless of the number of the level 2 units,the Bayesian estimate of random effects residual variance is more precise than PQL and the PQL estimation deviates from the true value seriously; In terms of the fixed effects estimates, When the number of level 2 units is 20, The mean and median of the Bayesian estimate is more precise than PQL,But With the number of level 2 units increase,the estimate of two methods is similar. Thus, in practical applications, we recommend the use of Bayesian method.

第二章针对不同高水平单位数，低水平数较小且不平衡的GLMMs数据，进行模拟研究，研究表明：无论水平2单位数的大小，贝叶斯方法随机效应残差方差估计值远比PQL法更接近真实值，PQL法严重偏离了真实值；而对固定效应参数估计值，当水平2单位数为20时，贝叶斯法各参数的均数和中位数均要较PQL法更接近真实值，而随着水平2单位数的增加，两种方法在估计固定效应参数时相差并不大，因而，在实际应用中，推荐采用贝叶斯法。

Then we introduced transition probability kernel for the hidden state process and the confederated process. In the deep research, we found that some similar parameters of state transition probability, observation symbol probability and initial state probability were involved in the transition probability kernel of the confederated process. So we can complete the training of traditional HMM parameters by training the transition probability kernel of the confederated process. To our surprised, we got the consistency of transition probability kernel of the confederated process. Meanwhile, we proved the rationality and reliability of the parameters estimation equations under the classical B-W algorithm.

然后对联合过程的转移概率核展开研究，我们根据过程的取值空间的不同，引入了隐状态空间和联合过程空间上的转移概率核，研究中发现在联合空间上的转移概率核包含了在一般的隐马氏模型的参数估计中的所有类似的三组参数(转移概率，观测概率和初始概率)，因此对传统的隐马氏模型的参数训练我们可以通过联合过程的转移概率核的训练来完成，更令我们惊奇的是联合过程的转移概率核还具有一致性；同时对于Baum提出的经典的B-W算法给出的关于模型的参数重估计公式，证明了它的合理性和可靠性，即模型经过重估计后的参数能够满足给定的单观测序列在模型下发生的概率是单调递增的。

The method of obtaining time delay parameters with the reverse access channel is discussed. The modified algorithms of classic correlation methods of TOA estimation are proposed. Noncoherently combining algorithm and group coherent combining algorithm are given. They decrease the computational complexity and are more suitable for implementations in practical systems. The leading edge detection algorithm is proposed to mitigate the effect of multipath in TOA estimation. Then the secondary search method is proposed, it can solve the matching issue of PN code in CDMA systems. The location algorithms based on LS are discussed; include circular trilateration with LS, hyperbolic trilateration with reference differencing, hyperbolic trilateration with sequential differencing and circular trilateration with TOA differences. The design and implementation of wireless location system based on TOA/TDOA in CDMA2000 systems are given. The location process is discussed, and the hardware and software implementation of TOA estimation are given. Finally, the wireless location system which uses TOA estimation and LS location algorithms above is tested in the real environment.

文中讨论了采用CDMA2000反向链路中的接入信道来获取时延的方法和过程；对TOA估计的传统的相关法进行了改进，提出了非相干合并相关法和分组合并相关法，减少了基于相关的时延估计算法的运算量，提高了算法的实用性，对算法进行了仿真和性能分析；采用边缘检测技术有效减少了TOA参数估计中的多径干扰，提高了时延参数估计的准确度；提出了一种TOA估计的二次搜索法，解决了多个定位测量单元同时进行TOA估计的PN码匹配问题，进一步了提高TOA估计的精度；在采用时延值的定位算法中引入了基于LS的圆周法、基于参考差的双曲线法、基于相邻差的双曲线法和基于TDOA的圆周法，讨论了各算法的应用条件，进行了算法性能的仿真分析；给出了CDMA2000系统中的TOA/TDOA的无线定位的系统方案和实现流程，并对TOA值估计算法进行了软硬件实现，还对采用TOA估计算法、LS定位算法和定位系统设计方案实现的定位系统的进行了现场定位性能测试。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。