生存數(shù)據(jù)的密度函數(shù)估計(jì)問題.pdf

1、北京化工大學(xué)碩士學(xué)位論文生存數(shù)據(jù)的密度函數(shù)估計(jì)問題姓名：RobertS.Toweh，BSc申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：應(yīng)用數(shù)學(xué)指導(dǎo)教師：施小丁20100526英文摘要TheProblemofdensityestimationofsurvivaldataAbstract：SurvivalanalyzehasbeendevelopedasanewbranchofstatisticsforrecentlytwentyyearsItisbrough

2、toutbytheneedofmanypracticeproblems，suchasmedicine，lifescience，reliabilityscienceandinsurance，etcItcanberegardedasthestatistictechniqueofanalysislifetimedataSurvivalanalyzesfocusesonthestatisticanalysisofcensoreddataRigh

3、trandomcensorshipisthemostimportantstatisticdataintermsoflifescience，medicinalpursuing，survivaltestofreliabilityandotherpracticalproblems，whoseresearchmethodsandtheoriesnotonlycouldbeenappliedintolifescience，publichealth

4、，reliabilityengineer，butalsodealwiththeproblemsofinsurancemathematics，criminologysociologycircumstancescience，academicaviationsubjectMeanwhile，thissubjecthasaextensivelyappliedforegroundAtfirst，thispaperdiscussedkernelde

5、nsityestimation，whichdealswithuncensoreddataThenweusesimulativemethodtocomparedifferentkernelfunctionandbandwidths’results，inordertoselectbetterkernelstocalculate—Epanechnikovkernel。Whatismore，wealsosimplydiscussedthesim

6、ulationeffectsofdifferentkindsofbandwidthsAtthelastofthispart，theproblemsofn—dimensionkernelestimationhadbeenresearched，andwesimulatedtwodimensionsituationSecondlyunderthecircumstanceofcensoreddata，wecomparedthedifferenc

7、ebetweenestimativemethodofsurvivaldataandnon—parameterkernelestimation，whichaccountedforwhythefirstmethodisnecessaryforestimatingsurvivaldataFarthermore，KaplanMeierproduct—limitestimatorandothermethodsgotinvolvedinthispa

8、perFinallythemaindifficultyofkerneldensityestimationsinpracticeliesinselectingallappropriatevalueofthesmoothingparameterthebandwidth，forusewithagivendatasetInthispaperweconsiderthebandwidthselectionproblemforkerneldensit

9、yestimationfromrandomlyfightcensoreddatabasedonameanabsoluteerrorcriterion1astbutnottheleast，whatistheinnovationofthispaperwesynthesizedtwomethodstoperformbetterespeciallyonboundaryKeywords：KernelEstimation；Randomcensore