Statisticalmethods
Statisticsarerichandintricate,anditisnoteasytoselectstatisticalanalysismethodsreasonably.Forthesamedata,ifyouchoosedifferentstatisticalanalysismethods,sometimestheconclusionsarecompletelydifferent.
Thebasisforthecorrectchoiceofstatisticalmethodis:
①Accordingtothepurposeoftheresearch,clarifythetypeofresearchexperimentdesign,researchfactorsandlevelnumber;
②DeterminethedataCharacteristics(whethernormaldistribution,etc.)andsamplesize;
③Correctlyjudgethetypeofstatisticaldata(measurement,count,andgradedata),andperformcorrectstatisticsaccordingtotheappropriateconditionsofthestatisticalmethodMeasurevaluecalculation;
Finally,accordingtotheactualsituationofprofessionalknowledgeanddata,combinedwithstatisticalprinciples,thestatisticalanalysismethodshouldbeflexiblyselected.
Statisticalclassification
1Statisticalmethodsofmeasurementdata
Statisticalanalysismethodsforanalyzingmeasurementdatacanbedividedintoparametrictestmethodsandnon-parametrictestmethods.
Parametertestingmethodsaremainlyt-testandanalysisofvariance(ANOVA,F-test),etc.Thet-testandu-testarecommonlyusedwhencomparingmeansbetweentwogroups,andtheanalysisofvarianceiscommonlyusedwhencomparingmeansabovetwogroups.;Non-parametrictestmethodsmainlyincluderanksumtestandsoon.Thettestcanbedividedintothettestofsinglegroupdesigndata,thettestofpaireddesigndataandthettestofgroupdesigndata;whentwosmallsamplesarecompared,thetwopopulationdistributionsarerequiredtobenormallydistributedandthevariancesarehomogeneous.IfnotTomeettheaboverequirements,anon-parametricmethod(ranksumtest)shouldbeused.Theanalysisofvariancecanbeusedtocomparethemeanoftwoormoresamples.Whenapplyingthismethod,eachsampleisrequiredtobeamutuallyindependentrandomsample,eachsamplecomesfromanormalpopulationandtheoverallvarianceofeachtreatmentgroupishomogeneous.Accordingtodifferentdesigntypes,therearemanydifferentmethodsintheanalysisofvariance.Forquantitativedata,appropriatestatisticalanalysismethodsshouldbeselectedaccordingtothetypeofdesignused,theconditionsofthedataandthepurposeofanalysis,andt-testandone-wayanalysisofvarianceshouldnotbeblindlyapplied.
2Statisticalmethodsofcountingdata
Thestatisticalmethodsofcountingdataaremainlyusedfortheanalysisofthefour-gridtableandtheR×Ctable
test.Fourgridtabledata:use
testorutestforcomparisonbetweengroups.Ifthetestcannotbesatisfied:whenthecountdataisinapaireddesign,thefourgridtableobtainedisapairedfourgridtable,andthetestusedForformulasandcorrectionformulas,pleaserefertobooks.TheR×Ctablecanbedividedintofourcategories:two-waydisorder,one-wayordered,two-wayorderedwiththesameattributes,andtwo-wayorderedwithdifferentattributes.Differenttypesofrowlistshavedifferentselectionmethodsaccordingtotheirresearchpurposes.SeedetailsTable1.
3Thestatisticalmethodofgradedata
Thegradedata(ordinalvariable)isthedataobtainedbygroupingthegradesofthenatureandcategory,andthencountingthenumberofobservationunitsineachgroup.Inclinicalmedicaldata,somequalitativeindicatorsareoftenencountered,suchasevaluationofclinicalefficacy,clinicalstageofdisease,clinicalgradingofdiseaseseverity,etc.Theseindicatorsareoftendividedintoseveralgradesandthenclassifiedandcountedtosolvetheirproblems.Quantitativeproblems,suchdataisstatisticallycalledgradedata.
ReferenceMaterials
[1]SuYinfa.Analysisoftherelationshipamongthethreestatisticalmethodsofhierarchicaldata[J].ChineseClinicalPharmacologyandTherapeutics,2002.7(2):175-176.
[2]XiaoMingyao,"ChinaMetrology"1996