differencedrewrite规则

Titlestata.
comarimapostestimation—PostestimationtoolsforarimaDescriptionSyntaxforpredictMenuforpredictOptionsforpredictRemarksandexamplesReferenceAlsoseeDescriptionThefollowingpostestimationcommandsareofspecialinterestafterarima:CommandDescriptionestatacplotestimateautocorrelationsandautocovariancesestatarootscheckstabilityconditionofestimatesirfcreateandanalyzeIRFspsdensityestimatethespectraldensityThefollowingstandardpostestimationcommandsarealsoavailable:CommandDescriptionestaticAkaike'sandSchwarz'sBayesianinformationcriteria(AICandBIC)estatsummarizesummarystatisticsfortheestimationsampleestatvcevariance–covariancematrixoftheestimators(VCE)estimatescatalogingestimationresultsforecastdynamicforecastsandsimulationslincompointestimates,standarderrors,testing,andinferenceforlinearcombinationsofcoefcientslrtestlikelihood-ratiotestmarginsmarginalmeans,predictivemargins,marginaleffects,andaveragemarginaleffectsmarginsplotgraphtheresultsfrommargins(proleplots,interactionplots,etc.
)nlcompointestimates,standarderrors,testing,andinferencefornonlinearcombinationsofcoefcientspredictpredictions,residuals,inuencestatistics,andotherdiagnosticmeasurespredictnlpointestimates,standarderrors,testing,andinferenceforgeneralizedpredictionstestWaldtestsofsimpleandcompositelinearhypothesestestnlWaldtestsofnonlinearhypotheses12arimapostestimation—PostestimationtoolsforarimaSyntaxforpredictpredicttypenewvarifin,statisticoptionsstatisticDescriptionMainxbpredictedvaluesformeanequation—thedifferencedseries;thedefaultstdpstandarderrorofthelinearpredictionypredictedvaluesforthemeanequationiny—theundifferencedseriesmsemeansquarederrorofthepredictedvaluesresidualsresidualsorpredictedinnovationsyresidualsresidualsorpredictedinnovationsiny,reversinganytime-seriesoperatorsThesestatisticsareavailablebothinandoutofsample;typepredict.
.
.
ife(sample).
.
.
ifwantedonlyfortheestimationsample.
PredictionsarenotavailableforconditionalARIMAmodelsttopaneldata.
optionsDescriptionOptionsdynamic(timeconstant)howtohandlethelagsofytt0(timeconstant)setstartingpointfortherecursionstotimeconstantstructuralcalculateconsideringthestructuralcomponentonlytimeconstantisa#oratimeliteral,suchastd(1jan1995)ortq(1995q1);seeConvenientlytypingSIFvaluesin[D]datetime.
MenuforpredictStatistics>Postestimation>Predictions,residuals,etc.
OptionsforpredictFivestatisticscanbecomputedusingpredictafterarima:thepredictionsfromthemodel(thedefaultalsogivenbyxb),thepredictionsafterreversinganytime-seriesoperatorsappliedtothedependentvariable(y),theMSEofxb(mse),thepredictionsofresidualsorinnovations(residual),andthepredictedresidualsorinnovationsintermsofy(yresiduals).
GiventhedynamicnatureoftheARMAcomponentandbecausethedependentvariablemightbedifferenced,thereareotherwaysofcomputingeach.
Wecanuseallthedataonthedependentvariablethatisavailablerightuptothetimeofeachprediction(thedefault,whichisoftencalledaone-stepprediction),orwecanusethedatauptoaparticulartime,afterwhichthepredictedvalueofthedependentvariableisusedrecursivelytomakelaterpredictions(dynamic()).
Eitherway,wecanconsiderorignoretheARMAdisturbancecomponent(thecomponentisconsideredbydefaultandisignoredifyouspecifystructural).
Allcalculationscanbemadeinoroutofsample.
arimapostestimation—Postestimationtoolsforarima3Mainxb,thedefault,calculatesthepredictionsfromthemodel.
IfD.
depvaristhedependentvariable,thesepredictionsareofD.
depvarandnotofdepvaritself.
stdpcalculatesthestandarderrorofthelinearpredictionxb.
stdpdoesnotincludethevariationarisingfromthedisturbanceequation;usemsetocalculatestandarderrorsandcondencebandsaroundthepredictedvalues.
yspeciesthatpredictionsofdepvarbemade,evenifthemodelwasspeciedintermsof,say,D.
depvar.
msecalculatestheMSEofthepredictions.
residualscalculatestheresiduals.
Ifnootheroptionsarespecied,thesearethepredictedinnovationst;thatis,theyincludetheARMAcomponent.
Ifstructuralisspecied,thesearetheresidualstfromthestructuralequation;seestructuralbelow.
yresidualscalculatestheresidualsintermsofdepvar,evenifthemodelwasspeciedintermsof,say,D.
depvar.
Aswithresiduals,theyresidualsarecomputedfromthemodel,includinganyARMAcomponent.
Ifstructuralisspecied,anyARMAcomponentisignored,andyresidualsaretheresidualsfromthestructuralequation;seestructuralbelow.
Optionsdynamic(timeconstant)specieshowlagsofytinthemodelaretobehandled.
Ifdynamic()isnotspecied,actualvaluesareusedeverywherethatlaggedvaluesofytappearinthemodeltoproduceone-step-aheadforecasts.
dynamic(timeconstant)producesdynamic(alsoknownasrecursive)forecasts.
timeconstantspecieswhentheforecastistoswitchfromonestepaheadtodynamic.
Indynamicforecasts,referencestoytevaluatetothepredictionofytforallperiodsatoraftertimeconstant;theyevaluatetotheactualvalueofytforallpriorperiods.
Forexample,dynamic(10)wouldcalculatepredictionsinwhichanyreferencetoytwitht00otherwisemeaningthatpredictnewvar,xbcalculatespredictionsbyusingthemetricofthedependentvariable.
Inthisexample,thedependentvariablerepresentedchangesinln(wpit),andsothepredictionsarelikewiseforchangesinthatvariable.
Ifweinsteaduse.
predicty,yStatacomputesytasyt=xbt+ln(wpit1)sothatytrepresentsthepredictedlevelsofln(wpit).
Ingeneral,predictnewvar,ywillreverseanytime-seriesoperatorsappliedtothedependentvariableduringestimation.
IfwewanttoignoretheARMAerrorcomponentswhenmakingpredictions,weusethestructuraloption,.
predictxbs,xbstructuralwhichgeneratesxbst=β0becausetherearenoregressorsinthismodel,and.
predictys,ystructuralgeneratesyst=β0+ln(wpit1)arimapostestimation—Postestimationtoolsforarima5Example1:DynamicforecastsAnattractivefeatureofthearimacommandistheabilitytomakedynamicforecasts.
Inexample4of[TS]arima,wetthemodelconsumpt=β0+β1m2t+tt=ρt1+θt1+tFirst,weretthemodelbyusingdataupthroughtherstquarterof1978,andthenwewillevaluatetheone-step-aheadanddynamicforecasts.
.
usehttp://www.
stata-press.
com/data/r13/friedman2.
keepiftimechi2=0.
0000OPGDS4.
lnm1Coef.
Std.
Err.
zP>|z|[95%Conf.
Interval]ARMAarL1.
.
3551862.
05030117.
060.
000.
2565979.
4537745L4.
-.
3275808.
0594953-5.
510.
000-.
4441895-.
210972/sigma.
0112678.
000488223.
080.
000.
0103109.
0122246Note:Thetestofthevarianceagainstzeroisonesided,andthetwo-sidedconfidenceintervalistruncatedatzero.
.
irfcreatenonseasonal,set(myirf)step(30)(filemyirf.
irfcreated)(filemyirf.
irfnowactive)(filemyirf.
irfupdated)WetthefollowingseasonalARIMAmodel(1ρ1L)(1ρ4,1L4)4lnm1t=tThecodebelowtsthisnonseasonalARIMAmodelandsavesasetofIRFresultstotheactiveIRFle,whichismyirf.
irf.
.
arimaDS4.
lnm1,ar(1)mar(1,4)noconstantnologARIMAregressionSample:1961q2-2008q2Numberofobs=189Waldchi2(2)=119.
78Loglikelihood=588.
6689Prob>chi2=0.
0000OPGDS4.
lnm1Coef.
Std.
Err.
zP>|z|[95%Conf.
Interval]ARMAarL1.
.
489277.
05380339.
090.
000.
3838245.
5947296ARMA4arL1.
-.
4688653.
0601248-7.
800.
000-.
5867076-.
3510229/sigma.
0107075.
000474722.
560.
000.
0097771.
0116379Note:Thetestofthevarianceagainstzeroisonesided,andthetwo-sidedconfidenceintervalistruncatedatzero.
.
irfcreateseasonal,step(30)(filemyirf.
irfupdated)8arimapostestimation—PostestimationtoolsforarimaWenowhavetwosetsofIRFresultsinthelemyirf.
irf.
WecangraphbothIRFfunctionssidebysidebycallingirfgraph.
.
irfgraphirfThetrajectoriesoftheIRFfunctionsaresimilar:eachgureshowsthatashocktolnm1causesatemporaryoscillationinlnm1thatdiesoutafterabout15timeperiods.
Thisbehaviorischaracteristicofshort-memoryprocesses.
See[TS]psdensityforanintroductiontoestimatingspectraldensitiesusingtheparametersestimatedbyarima.
ReferenceEnders,W.
2004.
AppliedEconometricTimeSeries.
2nded.
NewYork:Wiley.
Alsosee[TS]arima—ARIMA,ARMAX,andotherdynamicregressionmodels[TS]estatacplot—Plotparametricautocorrelationandautocovariancefunctions[TS]estataroots—CheckthestabilityconditionofARIMAestimates[TS]irf—CreateandanalyzeIRFs,dynamic-multiplierfunctions,andFEVDs[TS]psdensity—Parametricspectraldensityestimationafterarima,arma,anducm[U]20Estimationandpostestimationcommands