15SUOOUOU150U00U

108.1200001600000

1781300001700000

1883100001600000

iyU4bOQOU19OUOOUV

20<>

一圖

2.6.12.雙擊打開x序列表格形式，點擊表格左端View\Gragh\Line,或者在命

令框中輸入“linex”。可以看出X是有一定時間趨勢的，如圖2.6.2。

X

28

100200300400500600700

圖2.6.2可見序列x具有明顯的趨勢和季節波動，宜采用模型3或模型2檢驗。

3.點擊序列x表格上菜單命令：Vie叭UnitRootTest,出現對話框（圖2.6,3）

圖2.6.3可從圖中看一到，默認的檢驗方法為ADF,默認檢驗水平數據（原始數據，

后兩者1st、2nd為1階差分和二階差分數據），默認的檢驗模式為模型2。而

右邊在滯后階數的選取上，默認采用SIC最小。4.將檢驗模型改為模型3,其

余采用默認設定，點OK,出來結果如圖2.6.4：

Augmented(McKeyFuMerunitRootTestonX

NullH^otiesrsXhasaunitroot

ExogenousConstant

LagLengm0(AutomaticbasedonSIC.MAXL*?G=19)

t-StaksttcProb,

AugnientedDlcKey-FulleileslbtdUbVc-218293002128

Testaikcalvalues1%level-3439093

5%-286528g

10%level-2566822

,MadOnnon(1996)one-sidedHues

AugmentedDcKey+uiierlestkquanon

Depec&ntVariableD(X)

MethodLeastSquares

DaleW17/22Time1500

Sample(adiusted)2730

Includedobservations729adjstrnmU

圖2.6.4從結果可以判斷序列x有單位根。大家可以選擇其他模式和滯后期來檢

驗，以形成最終的判斷。檢驗序列X一階差分序列的平穩性：在圖2.6.3所示對

話框中選1stdiferent,檢驗模型為模型2,點OK,得下圖2.6.5

圖2.6.5從結果中可見序列X為一階差分平穩的，故序列x為一階單整的。

03建立ARIMA模型

趨勢圖：plotx或者linex一階差分之后的趨勢圖，顯示數據平穩

DX

先看自相關系數和偏相關系數圖Identx或者宜接點開

1

□Ip.1NTIT_ED::llntitlmd\1。II-II火1

ViewProcObjectPropertiesPrintNameFreezeSampleGenrSheetGraphStatsIdent

CorrelogramofD(X)

Date:08/17/22Time:17:23

Sample:1730

Includedobservations:729

AutocorrelationPartialCorrelationACPACQ-StatProb

1111-0.034-0.0340.84940.357

1'1120.0120.0110.94990.622

-1113-0.039-0.0392.09500.553

(14-0.080-0.0836.80460.147

1)50.0640.0609.83850.080

116-0.098-0.09516.9640.009

1'1170.0170.0C417.1870.016

'I||18-0.038-0.03818.2670.019

i11]90.0650.06621.4080.011

111110-0.016-0.03221.5990.017

1]1]110.0690.08125.1780.009

1111120.006-0.00325.2050.014

1||113-0.060-0.04527.8510.009

1)1||140.0550.04030.0930.007

'111115-0.041-0.01131.3540.008

)11160.0430.02232.7040.008

1'1||170.0170.03332.9110.012

1'1||180.0310.04033.6380.014

11119-0.002-0.01233.6420.020

111120-0.031-0.02134.3790.024

1121-0.005-0.01034.3970.033

-11122-0.039-0.02835.5540.034

1'11230.012-0.00335.6610.045

■1(1124-0.028-0.01236.2680.052

111125-0.017-0.03336.5000.064

1'11260.0250.01836.9890.075

31□270.1350.14150.8970.004

(1c128-0.055-0.07553.2070.003

)1]290.0630.07056.2490.002V

可嘗試AR——46927MA——4627

Isd(x)car(4)ar(6)ar(9)ar(27)ma(4)ma(6)ma(27)

DependentvmabtoD(X)

MefboaLeastSquares

Data08/-1tt22nm?:1022

Sample(aOuotedt29730

todudvdol??NVM*anM702AN

Coiwergenc*acr??v?。atttr105neratjons

MAgcast220

VanatteCocBaerrtStdError.WMcPtob

C00131W002Ml5061931006037

D03605060092431390139900001

AW)-04703260092032-a1104&400000

ARO)00303270025430154692701226

出(27)01931420051502374W7900002

也的0724/0.09bt>*)1bWM2/00000

MS)040$n50095367429$4>500000

MAO7)?00642250045MS-1)99011oie?i

R-squared0054039Meandependentvat00127M

A4USWOR-squared00444WSO3p?nd?ntvar0701904

SEofregression0686110Akaikeirfocriterion2095772

StMn$qu*o(11*13VflftQ7QStliaualtMton7147?M

L09tekahhood?727.6150HavunQumnafltr2115830

FsUks?c560^080DurMn-WatsonsUI289591

PfOD(F-fitaM?C)0000002

分析：F統計量顯示顯著，說明整個模型建立通過，DW值為2.0295,不存在自

相關性了。ar（9）sma：27）不顯著，可以考慮去掉，擬合優度僅0.054,很差，

人1（；值2.09（越小越好）。進一步，檢驗殘差是否為白噪聲序列，若為白噪聲序

列說明信息提取充分，否則需要重新建模。

subftyremCodoyESquaredKfduelg

HiUogrant?riormafityTeU

Ubei

Sen?lCcrre<?bOHIMT?t..

MM)0*>5T3SHetwkM圖困Tests.

-C0M225

RggrX0?M0399M8町00127M

Adiu^edRsquared00444908Cd€<)tndtnfver?70191M

S￡of06861103KBmfoeniefion2096772

$51scuMtdrtiid3X6379entwon21476?

U)gkit?8d?26159心MQurmenter21158W

5663680DuSWMsoe俎20A591

ProttiT?，啦"U08毆2

結果如下:

□i=1as

W>wProcObjectPrintNameFreueEsbmat*StatsRodi

~~((MTHnonNnat

DU*08/18J22Tim41029

Sampl929730

lnch>d?d。8”60ns7g

Q-3c?cprobaMt^stor7ARMAtorm(€)

AtMoconeiMonParvolCorreiaaonACPACOStaiProb

14O

13

2-0015W7015M

-0000-0.018M

3<KO2)4

s?00240Ot059G9

0027

562O1?11201

j?：??：43883

66O143

70.0150.O45449

，6

00140.4595?

80O17

0014482330028

9。^0?

0008O487230087

424

1"0?00210W8460159

8霹

M5.910044

v00800l2?/

00049MQ0M1

K)4或

O87

13-0080370025

1400340.4M1w40033

3藍

15<00020KO>0054

10O1515

0032120007

1700020.O150100

Z215993

1800130.O0137

191116123

00190.O3910174

201916黑

00200.M0214

210.O1?

<0003。O<l00273

22O5臼16s

Q(M71B0247

2M309.M37

00460226

4O19等

?OSH)?8700?S3

K4)M^”

■03。2Mu

"0O3/

0006K)0350

9^￡

00192121038

28V9要

29-004869<?220360

0.O24

300047-<>.ON60331

。24

31-0028>M3

gO425繪0355

?0031OQ270374

335Z<

乂-00510330

-<>O塔2fl

。4

0035O73144m30337

355a

%0063O篇02S6

0.17

0062O00222

-00.33

00090250

信息的提取并不充分，需要重新建模分析。

Isd(x)ca?4)ar(6)ar(27)ma(4)ma(6)

O|回

ViewProcObjectPrintNa'neFreezeEstimateForecastStatsResids

DependentVariable:D(X)

Method:LeastSquares

Date:08/18/22Time:11:07

Sample(adjusted):29733

Includedobservations:702afteradjustments

Convergenceachievedafter67iterations

MABackcast:2328

VariableCoefficientStd.Errort-StatisticProb.

C0.0130320.0246810.5280010.5977

AR(4)0.3157320.1077802.9294030.0035

AR(6)-0.4538910.109427-4.1478690.0000

AR(27)0.1362560.0321154.2427520.0000

MA(4)-0.4339290.109943-3.9468570.0001

MA(6)0.3885810.1102973.5230460.0005

R-squared0.049563Meandependentvar0.012764

AdjustedR-squared0.042735S.D.dependentvar0.701904

S.E.ofregression0.686742Akaikeinfocriterion2.094795

Sumsquaredresid328.2439Schwarzcriterion2.133717

Loglikelihood-729.2729Hannan-Quinncriter.2.109838

F-statistic7.258956Durbin-Watsonstat2.032869

Prob(F-statistic)0.000001

InvertedARRoots.92.90+.23i90-.23i85+.42i

3S-42i72-57i72+R7i54-72i

ODI回

ViewProcObjectPrintNa-neFreezeEstimateForecastStatsResids

CorrelogramofResiduals

Date:08/18/22Time:11:08

Sample:29730

Includedobservations:702

Q-statisticprobabilitiesadjustedfor5ARMAterm(s)

AutocorrelationPa巾alCorrelationACPACQ-StatProb

'I1111-0.017-0.0170.1943

12-0.004-0.0040.2033

1?I13-0.026-0.0260.6790

I11140.0210.0200.9962

I150.0730.0734.7266

I1116-0.014-0.0124.86210.027

I11170.0140.0155.00100.082

I11180.0060.0105.02580.170

I)190.0500.0476.83030.145

I11110-0.020-0.0227.11000.213

I]1]110.0780.08011.4310.076

111112-0.004-0.00111.4400.121

[1[113-0.066-0.07014.5360.069

1)11140.0440.04115.9380.068

111115-0.025-0.02316.3710.090

1111160.0280.01116.9500.109

1111170.0080.01616.9980.150

1111180.0240.02817.4250.181

?—A———C-A?——A—?v

依然較差。變換模型Isd(x)car(4)ar(6)ar(9)ar(ll)ar(27)ar(28)ar(29)

ma(4)ma(6)ma(ll)ma(27)

DependentVariable:D(X)

Method:LeastSquares

Date:08/18/22Time:11:21

Sample(adjusted):31730

Includedobservations:700afteradjustments

Convergenceachievedafter56iterations

MABackcast:430

VariableCoefficientStd.Errort-StalisticProb.

C0.0139470.0281940.4946740.6210

AR⑷0.4217270.0689616.11MOO0.0000

AR(6)-0.4412600.067856-6.5029070.0000

AR(9)0.0664130.0251552.64C1610.0085

AR(11)-0.1421360.058996-2.40S2570.0162

AR(27)0.3688030.0565716.51S3230.0000

AR(28)-0.0175390.022034-0.79E0160.4263

AR(29)0.0487280.0264431.8427230.0658

MA(4)-0.5493000.070791-7.75S5070.0000

MA(6)0.3691470.0692255.3325450.0000

MA(11)0.1843200.0585073.15C4100.0017

MA(27)-0.2532790.058224-4.35C0660.0000

R-squared0.069196Meandependentvar0.013129

AdjustedR-squared0.054314S.D.dependentvar0.702804

S.E.ofregression0.683452Akaikeinfocriterion2.093673

Sumsquaredresid321.3693Schwarzcriterion2.171692

Loglikelihood-720.7857Hannan-Quinncriter.2.123832

F-statistic4.649626Durbin-Watsonstat2.025204

Prob(F-statistic)0.000001

1-~ACC/人3Ancnc?cc;nccc;

經過反復嘗試,建立以下模型：lsd(x)car(4)ar(5)ar(6)ar(9)ar(11)ar(27)ar(28)

ar(29)ar(32)ma(4)ma(5)ma(6)ma(9)ma(11)ma(14)ma(27)ma(33)ma(34)

oTITLEDWorkfile:時間序列分析及預測案例…。II回次

ViewProcObjectPrintNameFreezeEstimateForecastStatsResids

A

DependentVariable:DQ()

Method:LeastSquares1

Date:08/18/22Time:13:08

Sample(adjusted):34733

Includedobservations:700afteradjustments

Convergenceachievedafter10iterations

MABackcast:033

VariableCoefficientStd.Errort-StatisticProb.

C0.0136300.0271680.5016690.6161

AR(4)-0.0281470.069105-0.4073030.6839

AR(5)-0.0767390.079579-0.9643170.3352

AR(6)-0.1866710.082358-2.2665930.0237

AR(9)0.4739810.0685366958430.0000

AR(11)-0.1269610.085173-1.4906170.1365

AR(27)-0.1962130.074157-2.6459300.0083

AR(28)-0.0386930.033844-1.1432640.2533

AR(29)0.0592430.0358571.6521750.0990

AR(32)-0.0271110.039327-0.6893610.4908

MA(4)-0.0791830.062794-1.2610050.2077

MA(5)0.1542530.0758712.0331010.0424

MA(6)0.1145170.0814001.4068440.1599

MA(9)-0.4659210.064482-7.2255600.0000

MA(11)0.2137020.0759392.8M1400.0050

MA(14)0.0405010.0343481.1791620.2387

MA(27)0.3426760.0694224.9361040.0000

MA(33)-0.1505720.042064-3.5795860.0004

MA(34)0.0638140.0315302.0238850.0434

R-squared0.109122Meandependentvar0.014214

AdjustedR-squared0.085575S.D.dependentvar0.702887

S.E.ofregression0.672140Akaikeinfocriterion2.070067

Sumsquaredresid307.6566Schwarzcriterion2.193596

Loglikelihood-705.5233Hannan-Quinnenter,2.117818

F-statistic4.634158Durbin-Watsonstat2.003860

Prob(F-statistic)0.000000

V

OEquation:UNTITLEDWorkfile:時間戶防吩析例1：：Unti.ill回I由^

ViewProcObjectPrintNai)eFreezeEstimateForecastStatsResids

CorrelogramofResiduals

Date:08/18/22Time:13:09A

Sample:34733

Includedobservations:700

Q-statisticprobabilitiesadjustedfor18ARMAterm(s)

AutocorrelationPartialCorrelationACPACQ-StatProb

I1111-0.002-0.0020.0027

11112-0.007-0.0070.0381

11113-0.019-0.0190.2875

111140.0120.0120.3913

11115-0.002-0.0020.3930

111160.0040.0030.4026

111170.0050.0060.4234

11118-0.031-0.0311.0957

11119-0.008-0.0081.1408

111110-0.000-0.0011.1408

1111110.0110.0101.2297

1111120.0080.0081.2706

1111113-0.025-0.0251.7204

1111140.0050.0051.7366

111115-0.003-0.0031.7418

1II1||160.0310.0292.4314

111117-0.000-0.0002.4315

1111180.0100.0092.4987

1111190.0100.0122.56450.109

(11120-0.054-0.0544.68440.096

111121-0.001-0.0024.68540.196

111122-0.012-0.0134.78390.310

1111230.0110.0094.87780.431

111124-0.035-0.0325.76880.450

111125-0.000-0.0005.76890.567

1111260.0030.0035.77500.672

1111270.0100.0095.84410.755

111128-0.005-0.0085.86290.827

1111290.0010.0025.86370.882

111130-0.027