学科分类
/ 1
3 个结果
  • 简介:因为为回归功能的有效推理要求heteroscedasticity应该被考虑,在回归分析检测heteroscedasticity的重要性广泛地被认出。在这篇论文,为heteroscedasticity的简单测试基于剩余分析在非参量的回归被建议。而且,有有Dette和Munk的比较的一些模拟是方法被进行评估建议测试的表演。结果证明在这篇论文的方法相当令人满意地表现并且在一些情况中是比Dette和Munk的方法更强大的。

  • 标签: 异方差性 非参数衰退 残差分析 非参数统计
  • 简介:Let{N(t),t≥0}bethenonhomogeneousPoissonprocesswithcumulativeintensityparameterΛ(t),{δt,t≥0}theageprocess,and{yi,t≥0)theresiduallifetimeprocess.Inthepresentpapertheexpressionsofn-dimensionalsurvivaldistributionfunctionsoftheprocesses{δi}and{yi},andtheirLebesguedecompositionsarederived.

  • 标签: 不均匀布松过程 剩余时间 勒贝格分解 分布函数
  • 简介:TheIncompleteOrthogonalizationMethod(IOM(q)),atruncatedversionoftheFullOrthogonalizationMethod(FOM)proposedbySaad,hasbeenusedforsolvinglargeunsymmetriclinearsystmes.However,theIOM(q)exhibitesirregularconvergencebehaviorwithwildoscillationintheresidualnormsthoughittendstodecreaseinaveryslowmanner,whichisowingtothelackofminimizationpropertyovertheKrylovsubspace.QMRmethodproposedbyFreund,GutknechtandNachtigal,owingtoitsabilitytoavoidbreakdownsandsmoothconvergencebehavior,isarobustiterativesolverforgeneralnonsingularunsymmetriclinearsystems.Inthispaper,weproposeanovelquasi-minimalresidual(QMR)variantoftheIncompleteOrthogonalizationMethod(IOM(q)).Numericalexpermentsshowthatithassmoothconvergencebehaviorandismoreeffective,especiallywhenusingitsrestartedversion.

  • 标签: unsymmetric linear systems IOM(q) Q MR