提出一种基于偏差信息准则(deriance information criterion,DIC)的鲁棒贝叶斯混合分布模型选择算法.在变分逼近框架下,给出鲁棒贝叶斯混合模型的DIC计算公式;设计的模型选择算法能同时学习模型参数推断和进行模型选择,避免在大的候选...提出一种基于偏差信息准则(deriance information criterion,DIC)的鲁棒贝叶斯混合分布模型选择算法.在变分逼近框架下,给出鲁棒贝叶斯混合模型的DIC计算公式;设计的模型选择算法能同时学习模型参数推断和进行模型选择,避免在大的候选模型集中根据模型选择准则选取最优模型.给出试验参数初始值设置方法,在含有较多离群点的仿真数据和Old Faithful Geyser数据上的试验结果表明了好的性能:得到鲁棒的混合分量参数和较准确的混合分量个数.展开更多
The presently existing decision making method for problem of goal type, i.e. the goal programming, is popular to some extent. In this paper we analyzed the features of the problem and the method,based on which we foun...The presently existing decision making method for problem of goal type, i.e. the goal programming, is popular to some extent. In this paper we analyzed the features of the problem and the method,based on which we found some defects of the method and pointed out these defects. To overcome these defects we absorbed the spirit and exploited concepts of evaluation criterion and the fault measure of evaluation criterion. We proposed and applied a method with an evaluation criterion, after which we also p...展开更多
A class of interactive multi objective decision making method by means of evaluation criterion is proposed for problems with linear value function,in which case,the decision maker(DM) usually has only unwhole infor...A class of interactive multi objective decision making method by means of evaluation criterion is proposed for problems with linear value function,in which case,the decision maker(DM) usually has only unwhole information of weights for objectives. The concept of fault measure of the evaluation criterion is proposed to measure the deviation of the evaluation criterion from the DMs preference structure.The approach to obtain an upper boundary of fault measure of an evaluation criterion,and the approach to modify the evaluation criterion to be one with smaller fault measure,and the approach to obtain a pre optimized objective set by evaluation criterion with certain fault measure are also proposed.展开更多
The objective of this study is to examine the effect of accounting standards and investor protection on value relevance of earnings and book value of equity among European Union countries during the years 1999-2007. T...The objective of this study is to examine the effect of accounting standards and investor protection on value relevance of earnings and book value of equity among European Union countries during the years 1999-2007. The results indicate that the adoption of International Financial Reporting Standards [IFRS] leads to improvement in value relevance, particularly on earnings. We also examine the impact of investor protection and the deviation of local accounting standards from IFRS on the effectiveness of the IFRS adoption. The results show supporting evidence for investor protection but inconclusive evidence for accounting standard deviation. However, additional analysis indicates that the countries which apparently benefit from adopting IFRS are those with high deviation of local accounting standards from IFRS and high investor protection. The findings imply that adopting IFRS alone cannot improve value relevance of accounting information, but standard setters and regulators need to strengthen their investor protection mechanisms in order to improve the quality of accounting information.展开更多
针对国际学生评估项目2015年数据(PISA),采用单参数、双参数和三参数的项目反应模型进行建模,在每个模型下,分别采用logistic连接函数和probit连接函数。针对6个模型,应用偏差信息准则(Deviance Information Criterion,DIC)和伪边际似...针对国际学生评估项目2015年数据(PISA),采用单参数、双参数和三参数的项目反应模型进行建模,在每个模型下,分别采用logistic连接函数和probit连接函数。针对6个模型,应用偏差信息准则(Deviance Information Criterion,DIC)和伪边际似然对数(Logarithm of Pseudo-Marginal Likelihood,LPML)进行模型评价和模型选择。结果表明,当连接函数为logistic双参数的项目反应模型表现最好,因为这个模型下的DIC值最小,并且LPML值最大。我们采用R软件nimble包进行编程。展开更多
文摘提出一种基于偏差信息准则(deriance information criterion,DIC)的鲁棒贝叶斯混合分布模型选择算法.在变分逼近框架下,给出鲁棒贝叶斯混合模型的DIC计算公式;设计的模型选择算法能同时学习模型参数推断和进行模型选择,避免在大的候选模型集中根据模型选择准则选取最优模型.给出试验参数初始值设置方法,在含有较多离群点的仿真数据和Old Faithful Geyser数据上的试验结果表明了好的性能:得到鲁棒的混合分量参数和较准确的混合分量个数.
文摘The presently existing decision making method for problem of goal type, i.e. the goal programming, is popular to some extent. In this paper we analyzed the features of the problem and the method,based on which we found some defects of the method and pointed out these defects. To overcome these defects we absorbed the spirit and exploited concepts of evaluation criterion and the fault measure of evaluation criterion. We proposed and applied a method with an evaluation criterion, after which we also p...
文摘A class of interactive multi objective decision making method by means of evaluation criterion is proposed for problems with linear value function,in which case,the decision maker(DM) usually has only unwhole information of weights for objectives. The concept of fault measure of the evaluation criterion is proposed to measure the deviation of the evaluation criterion from the DMs preference structure.The approach to obtain an upper boundary of fault measure of an evaluation criterion,and the approach to modify the evaluation criterion to be one with smaller fault measure,and the approach to obtain a pre optimized objective set by evaluation criterion with certain fault measure are also proposed.
文摘The objective of this study is to examine the effect of accounting standards and investor protection on value relevance of earnings and book value of equity among European Union countries during the years 1999-2007. The results indicate that the adoption of International Financial Reporting Standards [IFRS] leads to improvement in value relevance, particularly on earnings. We also examine the impact of investor protection and the deviation of local accounting standards from IFRS on the effectiveness of the IFRS adoption. The results show supporting evidence for investor protection but inconclusive evidence for accounting standard deviation. However, additional analysis indicates that the countries which apparently benefit from adopting IFRS are those with high deviation of local accounting standards from IFRS and high investor protection. The findings imply that adopting IFRS alone cannot improve value relevance of accounting information, but standard setters and regulators need to strengthen their investor protection mechanisms in order to improve the quality of accounting information.
文摘针对国际学生评估项目2015年数据(PISA),采用单参数、双参数和三参数的项目反应模型进行建模,在每个模型下,分别采用logistic连接函数和probit连接函数。针对6个模型,应用偏差信息准则(Deviance Information Criterion,DIC)和伪边际似然对数(Logarithm of Pseudo-Marginal Likelihood,LPML)进行模型评价和模型选择。结果表明,当连接函数为logistic双参数的项目反应模型表现最好,因为这个模型下的DIC值最小,并且LPML值最大。我们采用R软件nimble包进行编程。
