1、coefficientofdeterminationThecoefficientofdeterminationistheratioofthesumofsquaredregressiontothesumofsquaredtotal.thecoefficientofdeterminationissuchthat0=<r2=<1,anddenotesthestrengthofthelinearassociationbetweenxandy.Thecoefficientofdeterminationrepresentsthepercentofthedatathatistheclosesttothelineofbestfit2、p-valueInstatisticalsignificancetesting,thep-valueistheprobabilityofobtainingateststatisticatleastasextremeastheonethatwasactuallyobserved,assumingthatthenullhypothesisistrue.Oneoften"rejectsthenullhypothesis"whenthep-valueislessthanthesignificancelevela,whichisoften0.05or0.01.Whenthenullhypothesisisrejected,theresultissaidtobestatisticallysignificant.在零假設(shè)成立條件下，出現(xiàn)統(tǒng)計(jì)量目前值及更不利于零假設(shè)數(shù)值的概率3、confidenceintervalatypeofstatisticalintervalestimateforanunknownparameter:arangeofvaluesbelievedtocontaintheparameter,withapredetermineddegreeofconfidence.Itsendpointsaretheconfidencelimits,andithasastatedprobability(confidencecoefficient)ofcontainingtheparameter.Therangeofnumericalvaluesinwhichonecanbeconfident(toacomputedprobability—e.g.,90or95%),thatthepopulationvaluebeingestimatedwillbefound.4、lifetableAstatisticalmodelformeasuringthemortality(oranyothertypeof“exit”)experiencesofapopulation,controllingforagedistributionsmaybeacurrentlifetable,whenallthepeopleinapopulationatonetimearesurveyed,oracohorttable,whenallofthepeopleborninaparticulartimespanaredealtwithasagroup.5、Meta-Analysisasystematicmethodofevaluatingstatisticaldatabasedonresultsofseveralindependentstudiesofthesameproblem.6、PartialregressioncoefficientStatisticsinthepopulationmultiplelinearregressionequationthatindicatetheeffectofeachindependentvariableonthedependentvariablewiththeinfluenceofalltheremainingvariablesheldconstant;eachcoefficientistheslopebetweenthedependentvariableandeachoftheindependentvariables.一、standarddeviationandStandarderrormeaning:ameasureofdispersionforindividualsaroundsamplemean.itcanbeusedasthepointestimationofpopulationstandarddeviation。usedinthedescriptiveanalysisameasureofdispersionofsamplemeansfrompopulationmeanitcanbeusedasanestimationofsamplemeans'standarddeviationusedininferentialanalysisapplication:ifthestandarddeviationissmaller,itmansvariablesaremorecloselyarounddistribution,sothemeanhaveabetter.RepresentativeifStandarderrorissmaller,itmeansthatthedifferencebetweensamplemeanandpopulationmeanissmaller,usingthesamplemeantoestimatethepopulationmeanhaveabiggerreliabilityrelationshipwithn:thebiggernis,themoreclosersamplestandarddeviationandpopulationstandarddeviationthebiggernis,thesmallerStandarderrorwillbe.RelationshipbetweenSD&SESEisdirectlyproportionaltoSD,anditisinverseproportiontothesquare-rootofsamplesizen.theyareusedasStatisticalparametertodescribethedegreeofvariation二、confidencecoefficientandreferencerangeMeaning:atypeofstatisticalintervalestimateforanunknownparameter:arangeofvaluesbelievedtocontaintheparameter,withapredetermineddegreeofconfidence.Itsendpointsaretheconfidencelimits,andithasastatedprobability(confidencecoefficient)ofcontainingtheparameter.Therangeofnumericalvaluesinwhichonecanbeconfident(toacomputedprobability—e.g.,90or95%),thatthepopulationvaluebeingestimatedwillbefound.Asetofvaluesestablishedasnormalmaximumsorminimumsforagivenanalyteitisafluctuatingrange,application:toestimatepopulationmean,ititfluctuationrangeofpopulationmeanitcanreflectthedistributionrangeofacertainindexofMost(95%)observationobjectSamplesize:iftheintervalisbigger,itwillbeapproachto0iftheintervalisbigger,itwillbemorestablecalculationformulaunknownnsmall:orunknownbutn>60［覽曲

'二：占：known亠宀Normaldistribution------SkewnessdistributionPX~P100X三、Linearcorrelationandregressionofdifferenceandassossiationsassossiations1foradatawhichicandobothrelatedanalysisandregressionanalysis,thepositivenegativenumberofrandbisthesame2、thehypothesistestingofrandbvalueisthesametheirtvaluesareequal3Thesamedata‘scorrelationcoefficientandregressioncoefficientcanmutualconversion4、useregressiontoexplaincorrelation,riscloserto1,explainstheyaremorerelevantdifference：1、Datasource:correlationanalysisisrequiredthatbothxandyfollownormaldistribution;butforsimplelinearregression,onlyyisrequiredfollowingnormaldistribution.2.application:correlationanalysisisemployedtomeasuretheassociationbetweentworandomvariables(bothxandyaretreatedsymmetrically)simplelinearregressionisemployedtomeasurethechangeinyforx(xistheindependentvarible,yisthedependentvariable)3.risadimensionlessnumber,ithasnounitofmeasurement;butbhasitsunitwhichrelatetoy.4、Valuerange:-1=《r=《1,-00=《b=《+00四、Parametersdeterminingsamplesizea:ProbabilityoftypeIerror(significancelevel),ifaissmaller,Needmoresamplesize3:ProbabilityoftypeIIerror(statisticalpower)3issmaller,TestpowerisbiggerNeedmoresamplesize?:Standarddeviationofpopulation.IfcrisbiggerNeedmoresamplesizeS:TolerabledifferencebetweentwopopulationmeansortwopopulationproportionsifSisbigger,Needlittlesamplesize五、2x2Tabletestmattersneedingattention1、InthetableEachgridshouldhasT>1,andthegridscontaining1<Tv5shouldnotexceed1/5ofthetotalgrids。Ifn>40,butthereis1<Tv5inthegrid,needtouseCorrectionformulaAlso,wecanIncreasesamplesizedeleteormergebyprofessionalknowledge、Switchtotwo-waydisorderRxCtableFisherexactprobabilitymethod2、afterrefusenullhypothesis,weshouldcarryoutComparisonbetweentwo3、Fortheclassificationoforderlyvariables,cannotusethegeneralchi-squaretestmethod六、Non-parametrictests1、Distributionunknown(conditionofparametricmethodsnotmet)2、Ordinaldata:datahavearankingbutnoclearnumericalinterpretation3、Non-precisedata(i.e:>80);Varianceobviouslynoteqaual4、Aquickandbriefanalysis(forpilotstudy).One-sideortwo-sideHasNoupperlimitNolowerlimit七、.StepsofhypothesistestHypothesistestingisrequiredtousesmallprobability,fromtheoppositeoftheproblem(H0)tojudgewhethertheproblem(H1)istrueindirectly.ThentheconditionwasestablishedinH0calculationofteststatistics,finallygetPvaluetojudge1、determineH0H1and2、Selecttestmethodandcalculateteststatistic,determinedistributionanddf.3、Findoutcriticalvalueandp-valuefromtableorcomputer.4、Makeinference:rejectH0ifp<notrejectH1ifp>.5Professionalconclusion:clinicalorpreventivemerits八、 cautioninuserelativemeasure1Thedenominatorshouldbebigenough!Otherwisetheabsolutemeasureshouldbeused.Example:Outof5cases,3werecured-60%?2Attentiontothepopulationwheretherelativemeasurecomesfrom.Prevalencerate:Populationisthestudentsinthesamegrade.Constitutes:Populationisallthepatients3、Pooledestimateofthefrequency:Pooledestimate=numerators/denominators4、Comparabilitybetweenfrequenciesorbetweenfrequencydistributions—Noticethebalanceofotherconditions5、Ifthedistributionsofothervariablesaredifferent,toimprovethecomparability,needed.6Tocomparetwosamples,hypothesistestisneeded.(SeeChisquaretest)九、t檢驗(yàn)與方差分析

“Standardization”is方差分析的基本思想是根據(jù)研究目的和設(shè)計(jì)類型， 將總變異中的離均差平方和 SS及其自由度「分別分解成相應(yīng)的若干部分， 然后求各相應(yīng)部分的變異； 再用各部分的變異與組內(nèi) (或誤差)變異進(jìn)行比較，得出統(tǒng)計(jì)量 F值；最后根據(jù)F值的大小確定P值，作出統(tǒng)計(jì)推斷。多于兩個(gè)變量間的均數(shù)比較要用方差分析方差分析的應(yīng)用條件為①各樣本須是相互獨(dú)立的隨機(jī)樣本；②各樣本來(lái)自正態(tài)分布總體；③各總體方差相等，即方差齊t檢驗(yàn)適用于兩個(gè)變量均數(shù)間的差異檢驗(yàn)， 。用于比較均值的t檢驗(yàn)可以分成三類，第一類的。后兩種設(shè)計(jì)類型的區(qū)別在于事先是否將兩組研究對(duì)象按照某一個(gè)或幾個(gè)方面的特征相似配成對(duì)子t檢驗(yàn)，都必須在滿足特定的前提條件下應(yīng)用才是合理的。t檢驗(yàn)的前提條件就是該組資料必須服從正態(tài)分布； 若是配對(duì)設(shè)計(jì)，每對(duì)數(shù)據(jù)的差值必須服從正態(tài)分布；若是成組設(shè)計(jì)，個(gè)體之間相互獨(dú)立，兩組資料均取自正態(tài)分布的總體，并滿足方差齊性。之所以需要這些前提條件， 是因?yàn)楸仨氃谶@樣的前提下所計(jì)算出的 t統(tǒng)計(jì)ttt十、pair-designedtestandtwoindependentsamplesttestTheso-calledpairedsamplereferstointhetwosamplesobservationobjectshavesomeconnectionorhavethesameimportantfeatures,thentheyarepaired.twoindividualofeachgrouparerandomlyassignedtoaccepttwokindsofprocessionsDividethetestsubjectsintotwotreatmentgroupsrandomly,Eachgroupacceptsonetreatmenttheycomefromtwoindependentsamples,tomakeainfererencewhethertheirpopulationmeanareequal十——、whatistherelationshipbetweenBinomialdistribution,poissondistributionnormaldistribution1、ifX-B(n,n)whennisverybig,andnisverysmallthenXfollowspoissondistributionapproximately?、ifX-B(n,