CAMO软件TheUnscrambler教程(DataAnalysis)

DataAnalysisusingTheUnscrambler®ChangchunInstituteofAppliedChemistryRanjitViswanathanManager–Sales(AsiaPacific)[e-mail]#[phone]Background•ChangchunghasgiventwosetsofFiles–JCAMP-DXfiles,assumedtobeSpectrareadings.•Wewillconsidertheseas“Input”(X)variables.–ExcelSheet,withProtein%andFat%ofSoybeans•Wewillconsidertheseas“Output”(Y)variables.–ThereisoneextraJCAMP-DXreadingofS74,whichdoesnothaveProtein/Fatvalues.•Wewilltreatthisas“UnknownSample”,whereProtein%andFat%havetobepredicted,usingTheUnscrambler.•TheobjectiveistoPredicttheProtein/Fat%valuesofS74.OverallStepsthatwearefollowing1.EnsureDataConsistencyinJCAMP-DXandExcel2.ImportdataintoTheUnscrambler3.SettingupthedatainTheUnscramblerforanalysis4.PreliminaryAnalysisusingTheUnscrambler5.CreateaRegressionModel6.ImproveRegressionModel7.SaveRegressionModel8.PredictionusingTheUnscrambler9.MoreoptionsforPredictionEnsureDataConsistencyinJCAMP-DXandExcelRenamesomefilesinWindowsExplorer•SinceweareimportingdatafromtwodifferentsourcesintoTheUnscrambler,itisimportanttoensurethatallsamples/variablesareinthesameorderinbothsources.•Renamefiles“S1_0.dx”to“S01_0.dx”•ThisensuresthatallthefileswillbelistedinAscendingorderinTheUnscrambler.•Similarly,rename“S2_0.dx”,“S3_0.dx”“S4_0.dx”and“S6_0.dx”byaddinga0beforethedigit.Resultofrenaming•Theorderofthefiles,AFTERrenaming.AnalyzeExcelFile•Thereshouldbenomismatchinthedatafromthetwosources.Therefore,foralltheIRreadings,pleaseensurethatthereareProtein/Fat%valuesavailable.•Currently,thereisamismatchinthedata.•InExcel,thesampleS69doesnothaveaJCAMPreading.–DeletethissamplefromExcel.•InExplorer,thereisadditionalreadingofS74,whichisnotthereinExcel.–AddS74samplename,withblankvaluesofProtein%andFat%.–Thiswillbethe“UnknownSample”,whichwecanpredict.ResultaftercorrectingExcelsheetImportdataintoTheUnscramblerOpenTheUnscrambler•PleaseshutallfileswithinTheUnscrambler,toensurethatyougetablank,greyscreen.TheJCAMP-DXImportwindowofTheUnscrambler.•FileImportJCAMP-DX•Navigatetocorrectfolder•SelectALLfiles•Thisisthescreenyouwillget•PressOKResultofFileImportCreatetwoextravariablesinTheUnscrambler,fortheOutputVariables.•Selectthedata(rangeC2:D30).Oncethisrangeishighlighted,movethemousetothecorneroftheselection,untilitchangestoafoursidedarrow(theexcelcursorfor“moving”aselection).Thefollowingimageisarepresentationofthearrow.•Atthispoint,simplyleftclickonthemouse.IMPORTANT:KEEPITPRESSED.•MovetheselectiontoTheUnscramblericoninthetaskbar,untilTheUnscramblerbecomesactive,andtheninthegreyareaofTheUnscrambler,simplyreleasethemouse.•ThiswouldhavedraggedanddroppedtheDataintoTheUnscrambler.•Similarly,youcandraganddropthenamesoftheSamplesandVariablesintoTheUnscrambler.•ForVariablesNames,please“drop”itontotheVariableNamespaceofdatatableinTheUnscrambler.•ForSampleNames,please“drop”itontotheSampleNamespaceofdatatableinTheUnscrambler.•ForVariablesandSampleswhenyouare“dropping”itintoTheUnscrambler,pleasekeeptheCTRLkeypressed.Thatwill“copy”itintoTheUnscrambler.Otherwise,ifyoudonotpresstheCTRLkey,itwilldeleteitfromExcel,and“paste”itintoTheUnscrambler.DragandDropdatafromExceltoTheUnscramblerPleaseclickonthepicture,toviewVideo.TheResultsinTheUnscrambler(Editor)SettingupthedatainTheUnscramblerforanalysisEditSet•GotoModifyEditSet•ThisallowsyoutodefineVariableSetsandSampleSets,whichcanhelpmakeanalysiseasier.EditVariableWindow•DefineaVariableSetcalled“OutputVariables”,andselectVariables1-2.•Similarly,createVariableSet“Protein%”(Variable1),and“Fat%”(Variable2)ResultsofVariableSetdefinitionsCreateSampleSets•SelectSampleSetsfromthedropdownbox•PressAddCreatingSampleSets•CreateSampleSetcalled“KnownSamples”,(Samples1-28)•Similarly,createanothersampleset,called“UnknownSample”(Sample29)•PressOKTheUnscramblerEditorEnablealltoolbars•GotoViewToolbars•EnablealltoolbarsPreliminaryAnalysisusingTheUnscramblerSelectAllsamples•SelectALLsamples.Thiscanbedonebyclickingonthesamplenames,andselectingallthesamples.PlotsLine•Select“InfraredSpectr”asthevariablesetLinePlot•Ifrequired,youcanalsopreprocessdatabygoingbacktotheEditor,andchoose•ModifyTransform–BaselineCorrection–MSC–SpectroscopicTransformationsusingAPforSpectroscopy–etcMaximumvariationinthespectraseemstobeinthisregionShortcuttoLinePlotCreateaRegressionModelCreatingaRegressionModel•GotoTaskRegression•SelectPLS2Method•SelectCrossValidationMethod–SelectUncertaintyTest•SelectSamplesas“KnownSamples”•SelectXVariablesas“INFRAREDSPECTR”•SelectYVariablesas“OutputVariables”–ClickonWeights–ClickSelectAll•PressOKRegressionCalculations…•PressViewRegressionOverviewIfyoudonotseethesetoolbars,pleasegotoViewToolbars,andenablealltoolbarsThistoolbarhelpsinMarkingsamples/variables.RegressionCoefficientsofIRSpectrum.Asyoucansee,TheUnscramblerautomaticallymarksouttheSignificantvariables.ThisissimilartowhatweexpectedfromtheLinePlotoftheIRSpectrum,wheretherewasmaximumvariationonthehigherspectrumvalues.ResidualValidationVariance,perPC.Wecansee,thatafterPC5,thereisnosignificantdecreaseinresidualvariance.Thisvalueindicatestheoptima