在Eviews中對(duì)時(shí)間序列進(jìn)行預(yù)測的詳細(xì)步驟

1、在Eviews中對(duì)時(shí)間序列進(jìn)行預(yù)測的詳細(xì)步驟、輸入數(shù)據(jù)1.1 翻開Eviews6.0,根據(jù)如下圖翻開工作表創(chuàng)立框.器 EVieTsFileEdit Obj ect片w Frsc 賓ick O£tiQiis; TLndor HelpFVorkfile. .I今emEtv* ks.Trograirt7ert FileImportExpertPi- intPrint Setup.1.2 在右上角的data specfication框中輸入起止年份(start data和end data)Torkfile CreateWorkfile structure t/peDated - regula

2、r frequency vIrregular Dted and Panel worhfilejs may be made from Unstructured uwrkfiles by later speciFyiny date andjar otkier id ent fi er serigOKCancelote specific atioriFrequency; AnnualStart date：1的0End date: 2mN 回 roe 寫(optional)岫Pa簧：1.3 輸入數(shù)據(jù)：在輸入框中輸入data gdp 本文采用的數(shù)據(jù)為19902021年的GDP值.當(dāng)然,data后面可以輸

3、入任何你想要定義的“英文名字的 EViewsF_ila Edi i Object Vi ew Pz'o c Qul ck ions Window KeLp data gdip輸入data gdp后注意按回車鍵,彈出表格窗口后在其中輸入數(shù)據(jù) 也可復(fù)制進(jìn)去 數(shù)據(jù)：ctrl+v鍵、平穩(wěn)性檢驗(yàn)2.1在翻開的數(shù)據(jù)窗口中點(diǎn)擊 View - Correlogram (1)在彈出的窗口中直接點(diǎn) OK即可；C(rrt已 1 0匕Specif ica.Correlogram of®ooLags to include122.2自相關(guān)圖和偏相關(guān)圖進(jìn)行分析:EViers - G工dup; tKOUFD

4、l Torkfile; UHTITLEP;UntitLedJLI File Edi ObjEGt Vi ew froc £uiuk 口gti中ns 上indo* Help 丫懵科)皿.切目田Frm jwamg忻網(wǎng)啊崛m pl f )永比j(5間引沖或Date: 12117/14 Time: 11:55Sample: 1990 2021Included observations: 23AC PAC Ci-Stat ProbAutocorrelatl on P artlal C orrel Mion10.029 0.92917.9580.00IEL20.GS7 -0,DS729.7621

5、 nnn130.509 -0.02537.2050.000140.381 口.03441.6070 000150.253 -0 09343.6540 000160.145 -0.03244.362o.ono10.060 -0.02044.4930 00016-0.012 -0.04544.438o.aao19-0,Q74 40434J,7220.000110-0.126 -0 04245.4200 ODO11-0.171 -0 05546.9190.00012-0.212 -0.06049.1 770.000最簡單粗暴的方法就是看最右邊的 Prob值即P值,當(dāng)這列數(shù)據(jù)有多數(shù)都大于 0.05置信

6、水平時(shí)為白噪聲序列=序列是平穩(wěn)的.本文中GDP數(shù)據(jù)P值均小于 0.05,那么為非白噪聲.需對(duì)序列進(jìn)行差分.三、取一階差分gdp進(jìn)行一階差分,一階差分后3.1在輸入框中輸入第二列代碼,這代表將數(shù)據(jù) 的值命名為dgdp.按回車鍵德 EVicrsFile Edit Object View Froc 如 ick Ogtions Wi data gdp genrtjgdp=d(gdp)|3.2在dgdp數(shù)據(jù)的窗口中重復(fù)2.1的操作,對(duì)序列的平穩(wěn)性進(jìn)行檢驗(yàn)得到結(jié)果如下:三xA_) /if：*"''it-'.-.' Pru_ | C'b fi_tlnL |

7、匚卜- j 13 iuW | 三TbJ i5Lut | 口鳳" | "T if J| 3=il i uIhRange. 1&注通|£ |me 1工向區(qū)ISdkxcs; DGDPVorkfxId: UHIZILED:;Until1c dF- mtXliw Frocl口bjted Prcpotic (Pfnt NMneFrEZ七|三叩七J ?nr |'5h| S性目匹HE0clp uroupiCarrlo>grAiii of DGlDP3 nid12/14 T|lfiri&-11：56Sample； 1900 2021inclLdgd 口

8、口仁0gtluns 22A£itoc err t Inti o nPAO O-Qtal: Prob口 匚 1=g 101 1 112I 773I 530D.495D.42E-D.337D.1D5 .Di 1O.IOS 1必-D.223-D.254-0.7920 773-1.169J.377-1.324-0.137 -i I ：i-0.032-J.050ijsai o -1.037-U.0j415 025 Z2 42920.2 SO34 57136 31036 E7337 00538 621 40.81213,906414o u o o o O n- o c o o D o u o o

9、 o O n o o o- DD U DD 0 O PO OGO O.OW 0.0 CdO u oou慘！還是非白噪聲,只能進(jìn)行二階差分了!四、取二階差分4.1如第三列代碼所示記得不能重復(fù)命名ETievsFile Edit Object Fieh Froc Quick Options ffindov Heidata gdpgenr dgdp=d(gdp)genr dgdp2=d(gdp,2)| forkfile: USTITLEDViewllProc lobect| Printlfsave IlDetails+J- |5how|Fetch5tore |AL4 上kc LEi-Range: 1S

10、ample：!nn n 門口1/ _ _一 chm Series: DGDPTorkfile回CS gdp 目 group 必 residS dgdpf vp yWi炳總睥舊匚JObje=Sj Pr評(píng)曰tes Print | Name | Fr8知Date: 12/17/14 Time: 11:56Sample: 1990 201 2Con eloqi arn oT4.2對(duì)新的序列dgdp2進(jìn)行平穩(wěn)性檢驗(yàn),步驟同上,結(jié)果如下:MY GOD!看見了木有,這回是白噪聲了, P值多數(shù)都大于0.05!五、用最小二乘法對(duì)模型進(jìn)行估計(jì)：輸入ls dgdp2 c ar 2探索性建模5.1AR 2模型結(jié)果準(zhǔn)確

11、的說這個(gè)模型應(yīng)該是 ARIMA的疏系數(shù)模型,本文重點(diǎn)不在這！如有需要請(qǐng)私信我！恥1 Object 工i<v rec Juid 型后?虱m 1口gsnr期中她中二何區(qū)C|X1伽|際耳悔用pfht忸'而etL卜咂用Uh 5忙re 同ete原Display Filer*Rango: W9JX12 - 23 q幗 sample.J 990 2021 - 23 cbs Equation: EQU1 Torkfile: HITnLED:Enli. ITBIRIs如航行10inducedCeptnuemtUariatilE. DGDP2WdhDd: L的st湖幽匚# 12/1TJ1J Time

12、-11:57Sample (adiustec): 1994 2021nc u ded o bseiai on s. 19 &fter a j ustmenteComeigence achieved ater4 ileiaions工為淅聯(lián)MStlEriflf代廉;i:Prob.八JJ附led /C?453,煙1216.633 2.00171Q匚呻 5IARffl小潮出0216317-16753150.0019R-gquaiedD,M3I71Meai deperdBUtvar1997,664Adjusted RSQuaRdJ 410416B D. d epence nt vai12474

13、02SlE ofragiessiDnM75.055Aik?infoui1erion21.271651&jrn£Q3iM既M1 56EK9 S邪砒幅制21.3106bOMI(削hMd-20 ：HL6 Hchn-u rn(inH.212期1F-mti血13.53DOODurtiii岫twi 對(duì)13627UPr曲北聞.IDtilBEiD.5te： 135.2MA(2)模型結(jié)果5.3優(yōu)化模型：根據(jù)AIC和SBC準(zhǔn)那么選擇模型,值越小的擬合效果越好,本文 的選擇MA(2)模型.5.4 對(duì)模型進(jìn)行檢驗(yàn)：View-Residual Tests - Correlogram Q statist

14、ics檢驗(yàn)結(jié)果如下:P值大于0.05,為白噪聲序列,那么平穩(wěn).六、預(yù)測在模型輸出結(jié)果窗口那5.4圖那,菜單欄由Forecast進(jìn)行預(yù)測,注意步驟如下:6.1如要預(yù)測下一期的數(shù)值記得 先將數(shù)據(jù)范圍進(jìn)行修改在1.3那的窗口處在Range那對(duì)1990 2021雙擊；彈出窗口后進(jìn)行數(shù)據(jù)修改將2021改成2021 你想多預(yù)測也行,可以試試看壞笑,其實(shí)不能預(yù)測太多6.2 改了數(shù)據(jù)范圍后,去5.4的那個(gè)窗口點(diǎn)Forecast在彈出的窗體中,有個(gè)method 的框,默認(rèn)的是動(dòng)態(tài)的,你要點(diǎn) static ,然后點(diǎn)OK6.3 新生成的變量自動(dòng)命名為dgdp2f,這個(gè)名字只是一個(gè)名字！你就算把它改成gdp,它的數(shù)值還是二階差分后的數(shù)值！重點(diǎn)來了 ！如何根據(jù)二階差分后的數(shù)值計(jì)算原數(shù)值就是實(shí)際上你要預(yù)測的那個(gè)