非随机抽样和随机抽样的比较

Ch.7Selectingsamples1.Theneedtosample2.OverviewofSamplingtechniques3.Probabilitysampling4.Non-probabilitysamplingDefinitionofterms•CensusCollectandanalysedatafromeverypossiblecaseorgroupmember•SamplingArangeofmethodsthatenableresearchertoreducetheamountofdatabyonlydatafromasubgroupratherthanallpossiblecasesorelements•PopulationThefullsetofcasesfromwhichasampleistakenFigure7.1Population,sampleandindividualcases1.Theneedtosample•Budgetconstraintspreventyoufromsurveyingtheentirepopulation•Timeconstraintspreventyoufromsurveyingtheentirepopulation•Impracticabletosurveytheentirepopulation•Youhavecollectedallthedatabutneedtheresultsquickly2.Overviewofsamplingtechniques•Probabilityorrepresentativesampling-Eachcasefrompopulationisknownandusuallyisequalforallcases-(surveyandexperimentalresearchstrategies)•Non-probabilityorjudgementalsampling-Probabilityofeachcasefromthepopulationisunknown-Impossibletoanswerresearchquestionsortoaddressobjectivesthatrequireyoutomakestatisticalinferencesaboutthecharacteristicsofthepopulation-(casestudystrategy)非随机抽样和随机抽样的比较抽样方法作用抽样原则误差判断应用优缺点非随机抽样研究总体的局部现象非随机抽出样本，主观性强不能计算和判断抽样误差可随时随地采用不够科学规范，但省钱、省事、灵活方便随机抽样以部分推断总体随机抽出样本，客观性强不能计算和判断抽样误差只能定期采用科学规范，但费时、费钱、不够灵活方便Figure7.2Samplingtechniques随机抽样非随机抽样简单随机抽样系统抽样分层抽样分群抽样多步抽样配额抽样雪球抽样便利抽样自选抽样目的抽样极端抽样同质抽样不均匀抽样典型抽样关键抽样3.Probabilitysampling•Processofprobabilitysampling(2)decideonasuitablesamplesize(3)selectthemostappropriatesamplingtechniqueandselectthesample(4)checkthesampleisrepresentativeofthepopulation(1)identifyasuitablesamplingframebasedonyourresearchquestionsorobjectives•SampleframeAcompletelistofallthecasesinthepopulationfromwhichyoursamplewillbedrawn•Samplesizethenumberofcasesusedfortheresearchanalysis•Statisticalinferenceaprobableconclusionaboutapopulationonthebasisofdataofsample•LawoflargenumberLargersamplesizecanbetterrepresentthepopulationthanSmallersamplesizeHowtochoosethesamplesize?•Theconfidenceyouneedtohaveinyourdata:thelevelofcertaintythatthesamplecanrepresentthetotalpopulation(confidence↑samplesize↓)•Themarginoferrorthatyoucantolerate:theaccuracyyourequireforanyestimatesmadefromyoursample(accuracy↑samplesize↓)•Thetypesofanalysisyouwillundertake:(Categories↑samplesize↑;minimumthresholdofeachtechnique)•ThesizeoftotalpopulationfromwhichyoursampleisbeingdrawnResponserate•Reasonsofnon-response:Unreachable;ineligible;inability；refusal;totalnumberofresponses•TotalResponserate=----------------------------------totalnumberinsample-ineligibletotalnumberofresponses•ActiveResponserate=------------------------------------------totalnumberinsample–(ineligible+unreachable)Population,samplingframe,samplesSelectappropriatesamplingtechnique•Fivemainsamplingtechniques(1)simplerandom(2)systematic(3)stratifiedrandom(4)cluster(5)multi-stagesamplingtechnique(1)Simplerandomsampling(a)Numbereachcaseinyoursamplingframewithauniquenumber(b)Selectcasesusingrandomnumbersuntilyouractualsamplesizeisreached(pp218;587for“Randomnumbertables”).samplingtechnique(2)Systematicsampling(a)Numbereachcaseinyoursamplingframeatregularintervals(b)Selectthefirstcaseusingarandomnumber(c)calculatethesamplingfraction(抽样比)(d)selectsubsequentcasessystematicallyusingthesamplingfractiontodeterminethefrequencyofselectionactualsamplesize•Samplingfraction=-----------------------------totalpopulationSamplingfraction:Theproportionofthetotalpopulationthatyouneedtoselect.1.Decideonsamplesize:n2.DivideframeofNindividualsintongroupsofkindividuals:samplingfractionk=n/N3.Randomlyselectoneindividualfromthe1stgroup4.Selectevery1/k-thindividualthereafterSystematicSampleN=64n=81/k=8FirstGroupsamplingtechnique(3)Stratifiedrandomsampling[‘strætifaid](a)choosethestratificationvariable(s)(b)dividethesamplingframeintothediscretestrata(c)numbereachofthecaseswithineachstratumwithauniquenumber(d)selectyoursampleusingeithersimplerandomorsystematicsamplingStratifiedSample1.DividePopulationintoSubgroupsMutuallyExclusiveExhaustiveAtLeast1CommonCharacteristicofInterest2.SelectSimpleRandomSamplesfromSubgroupsAllStudentsPart-timeFull-timeSamplesamplingtechnique(4)Clustersampling(a)Choosetheclustergroupingforyoursamplingframe(b)numbereachoftheclusterswithauniquenumber.Thefirstclusterisnumbered0,thesecond1,andsoon(c)selectyoursampleusingsomeformofrandomsamplingClusterSample1.DividePopulationintoClustersIfManagersareElementsthenCompaniesareClusters2.RandomlySelectClusters3.SurveyAlloraRandomSampleofElementsinClusterCompanies(Clusters)Samplesamplingtechnique(5)Multi-stagesamplingOverviewofprobabilitysample1.Quotasampling2.Purposivesampling3.Snowballsampling4.Self-selectionsampling5.Conveniencesampling4.Non-probabilitysampling4.Non-probabilitysampling(1)Quotasampling(a)dividethepopulationintospecificgroups(b)calculateaquotaforeachgroupbasedonrelevantandavailabledata(c)giveeachintervieweranassignment,whichstatesthenumberofcasesineachquotafromwhichtheymustcollectdata(d)combinethedatacollectedbyinterviewertoprovidethefullsampleSamples4.Non-probabilitysampling（2）Purposivesampling(judgementalsampling)Useresearcher’sjudgemen