资源描述:
《the r book statistical modelling》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、9StatisticalModellingThehardestpartofanystatisticalworkisgettingstarted.Andoneofthehardestthingsaboutgettingstartedischoosingtherightkindofstatisticalanalysis.Thechoicedependsonthenatureofyourdataandontheparticularquestionyouaretryingtoanswer.Thekeyistou
2、nderstandwhatkindofresponsevariableyouhave,andtoknowthenatureofyourexplanatoryvariables.Theresponsevariableisthethingyouareworkingon:itisthevariablewhosevariationyouareattemptingtounderstand.Thisisthevariablethatgoesontheyaxisofthegraph.Theexplanatoryvar
3、iablegoesonthexaxisofthegraph;youareinterestedintheextenttowhichvariationintheresponsevariableisassociatedwithvariationintheexplanatoryvariable.Youalsoneedtoconsiderthewaythatthevariablesinyouranalysismeasurewhattheypurporttomeasure.Acontinuousmeasuremen
4、tisavariablesuchasheightorweightthatcantakeanyrealnumberedvalue.Acategoricalvariableisafactorwithtwoormorelevels:sexisafactorwithtwolevels(maleandfemale),andcolourmightbeafactorwithsevenlevels(red,orange,yellow,green,blue,indigo,violet).Itisessential,the
5、refore,thatyoucananswerthefollowingquestions:Whichofyourvariablesistheresponsevariable?Whicharetheexplanatoryvariables?Aretheexplanatoryvariablescontinuousorcategorical,oramixtureofboth?Whatkindofresponsevariabledoyouhave:isitacontinuousmeasurement,a
6、count,aproportion,atimeatdeath,oracategory?Thesesimplekeyswillleadyoutotheappropriatestatisticalmethod:Theexplanatoryvariables(a)AllexplanatoryvariablescontinuousRegression(b)AllexplanatoryvariablescategoricalAnalysisofvariance(ANOVA)(c)Explanatoryvariab
7、lesbothcontinuousandcategoricalAnalysisofcovariance(ANCOVA)TheRBook,SecondEdition.MichaelJ.Crawley.©2013JohnWiley&Sons,Ltd.Published2013byJohnWiley&Sons,Ltd.STATISTICALMODELLING389Theresponsevariable(a)ContinuousNormalregression,ANOVAorANCOVA(b)Proportio
8、nLogisticregression(c)CountLog-linearmodels(d)BinaryBinarylogisticanalysis(e)TimeatdeathSurvivalanalysisTheobjectistodeterminethevaluesoftheparametersinaspecificmodelthatleadtothebestfitofthemodeltothedata.Thedataaresacrosan