Overview
NumXLisasuiteoflightweighttimeseriesExceladdins.Ittransformsyour MicrosoftExcelapplicationintoafirstclasstimeseriesandeconometricstool, deliveringthekindofstatisticalaccuracyofferedbytheleadingstatisticspackages. NumXLintegratesnativelywithExcel,addingscoresofeconometricfunctions,arich setofshortcuts,andintuitiveuserinterfacestoguideyouthroughtheentireprocess. Whetheryouhaveasimplehomeworkproblemoralargescalebusinessproject, NumXLsimplifiesyourefforts.Ithelpsyoureachyourgoalsinthequickest,mostthoroughwaypossible. NumXLkeepsyourdataandresultsconnectedinExcel,lettingyoutraceyourcalculations,addnewdata pointsorupdateanexistinganalysis,easilysharingyourresultwithcoworkersand,yes,evenyour boss. Thelearningcurvecouldn'tbeeasier:NumXLrequiresnoprogrammingorscripting,andyouwontneed tomoveyourdatabetweenanyexternalprograms. Youcanalsodoanykindofadhocanalysis,asallofNumXLsfunctionsareaccessibleinyour spreadsheetandinsidetheVBAenvironment,shouldyouchoosetoscript.
What can NumXL do for me?
NumXLcomespackedwithscoresoffunctionsthatyoucaneasilyaccessthroughthefunctionwizardin Excel,aswellasseveralwizardsandshortcutUIstofacilitatethetimeseriesanalysisprocessand automatethemostcommonsteps(e.g.summarystats,modeling,calibration,diagnosis,forecast,and more.)
1. General Statistics
Usingthedescriptivestatisticsandcorrelogramwizards,youcanexaminethedataseriessummaryand timeseriesstatisticswithjustafewclicks.Thewizardscomewithanextensivesetofstatisticaltests, fromasimpleonesampletestformeantothemoresophisticatednormalityandARCHeffecttests. Furthermore, examiningtime dependency(auto correlation)in yoursampledata isonlyafewclicks away. Thewizards Overview 1 SpiderFinancialCorp,2012
generateprofessionallyorganizedtablesandgraphssummarizingallyourcalculations,readytobe includedinyourpresentation.Tomakethingseveneasier,alloutputsincludeNumXLfunctionsintheir formulaeforconnectingvalueswithinputs,soyoucanedit,updateorcustomizeasyouseefit.Youcan evenrerunthewizardifyoufeellazy. AsidefromtheWizards,NumXLhasnumerousfunctionsfor measuringpredictionpower(e.g.SSE,RMSE,MAPEand others). Tovisualizeyourdatadistribution,NumXLoffersseveral functionstoeasethetaskofconstructinghistograms, empiricalcumulativedistribution(EDF),andkerneldensity estimate(KDE).
2. Correlogram Analysis
Usingthecorrelogramwizardandfunction,constructingautocorrelationandpartialautocorrelation plotsisasnap. Furthermore,NumXLcomeswithsupportfor calculatingcrosscorrelationusingthree differentmethods:Pearson,Spearmanand Kendall. Finally,forcompletionpurposes,NumXL supportHurstexponentanalysisandGINI coefficient.
3. Statistical Testing
Forseriousdataanalysis,youmightneedtoconsiderthestatisticalsignificanceofacalculated parameter(e.g.autocorrelationfactor,excesskurtosis,etc.)orverifyanassumptionfromearlierdata (e.g.Normality,stationarity,absenceofserialcorrelationandothers).
NumXLpackagesthemajorityofthesetestsasasimpleAPItocalculatethePValueoftheunderlying test.
Overview
SpiderFinancialCorp,2012
4. Transformation
Datatransformationisacommonpreliminarystepinrealworldanalysisand/ormodeling.NumXL comeswithmostcommontransformationfunctions(e.g.BoxCox,differenceandseasonal difference/integraloperators,andothers.)
5. Smoothing
Smoothingandfilteringaretwoofthemost commonlyusedtimeseriestechniquesfor removingnoisefromtheunderlyingdatato helprevealtheimportantfeaturesand components(e.g.trend,seasonality,etc.). However,wecanalsousesmoothingtofillin missingvaluesand/orconductaforecast. NumXLsupportsseveralsmoothingfunctions,fromasimpleweightedmovingaverage(WMA)to Winterstripleexponentialsmoothingfunction. Ofcourse,wecanttalkaboutsmoothing withoutmentioningtrendfunctions.Trend analysisisveryoftenused(orabused)inthe industrytomakeaquick(anddirty)forecast. Executivesmightusethetrendingtoolasa sanitycheckwhenhe/sheexaminesresults frommoreadvancedmodels.NumXLsupports severalformsoftrend:linear,polynomial, power,exponentialandlogarithmic.
6. Calendar functionality
Calendareventsinfluencethevaluesofthetimeseriessample,andaprioradjustmentforthoseevents willhelpustobetterunderstandtheprocess,modelingandforecast. NumXLcomeswithnumerousfunctionstosupportcalendaradjustment,daterollingandadjustment, U.S.andnonU.Sbasedholidayssupport,nonwesternweekends,andpublicandbankholiday calendars.
7. Spectral Analysis
Instatistics,spectralanalysisisaprocedure thatdecomposesatimeseriesintoaspectrum ofcyclesofdifferentlengths.Spectralanalysis isalsoknownasfrequencydomainanalysis. NumXLcurrentlysupportthediscreteFourier Overview 3 SpiderFinancialCorp,2012
transformation(anditsinverse),withfutureplansformoreextensivecoverage.
8. ARMA/ARIMA
NumXLcomesloadedwith numerousfunctionstohelp youwithanyARMAanalysis task.Youcanstartby specifyingthemodelsorder usingtheARMAwizard.The wizardwilltieinmodel relatedcalculationsthe Loglikelihoodfunction, Akaikesinformationcriterion(AIC)andresidualdiagnosiswithyourinputdata.Oncethatisdone, fittingthemodelparameters(calibration)isasnap:simplyselectthemodelandclickon"calibration." Thesamegoesforforecasting:selectthemodeltableandclick"forecast". Youcanalwaysedittheformulasintheoutputcells,getintermediatecalculations(e.g.residuals,fitted mean)oruseaNumXLfunctioninyourownformulasorVBAcodeifyousochoose.
9. Seasonal ARIMA
NumXLsupportseasonalARIMAthroughtwodifferentmodels:(1)AirLine,and(2)X12ARIMA. X12ARIMAisthewidelyusedseasonal adjustmentprogramdeveloped,supportedand maintainedbyU.S.Censusbureau.NumXL providesanintuitiveinterfacewiththe programtohelpExcelusersmakeforecastsand seasonaladjustmentsquicklyandefficientlyfor economicandfinancialdata. NumXLalsooffersusersaccesstoallrawfiles(input/output)thataregeneratedintheprocessofthe dataanalysis.
10.
ARCH/GARCH
SimilartoARMA/ARIMA,modelingaGARCH typemodelisabreeze.UsingtheGARCH Wizard,youcangenerateamodeloutputtable withallcoefficientvaluesandrelated calculations(e.g.LLFandresidualdiagnosis). Thistablecanbeusedtocalibratethemodel andpredictoutofsamplevalues.
Overview
SpiderFinancialCorp,2012
NumXLalsosupportsARCH/GARCH,exponentialGARCH(EGARCH)andGARCHinmean(GARCHM) modelsnatively. Furthermore,NumXLsupportsGaussian,Student'standGEDtypeinnovations.
11.
ARMAGARCH mixture model
Intheeventyouwishtomodelthetimevaryingconditionalmeanandtheconditionalvolatilityinone model,anARMAGARCHmixturemodelisinorder.Bycombiningthetwomodels,theARMAwillfollow themeanandpasstheresidualstoGARCHtofollowthevarianceovertime. ARMAGARCHcombomodelssupportalltheARMAandGARCHmodelssupportedsolo,includingthe nonGaussianinnovations.
12.
Generalized Linear Model (GLM)
Whetheryouhavealogisticsregressionora generallinearmodelinmind,youcanusethe generalizedlinearmodel(GLM).NumXLusesthe GLMwizardtohelpyouspecifytheinputdata, linkfunctions(e.g.probit,logit,logcomplement) andgenerateamodeloutputtable. Aswithallofourmodels,theoutputmodel tableisusedtocalibrateandperformoutof sampleforecasts.
13.
Miscellaneous
Throughoutouranalysis,weimplementedafewfunctionsthatdidnotfitinwithanyofthoselisted above:interpolation,extrapolation,leastsquaresregressionandregularexpressionsfunctions.We combinedtheseorphanfunctionsintoasinglecategorycalledUtilities. Overview 5 SpiderFinancialCorp,2012