使用Eviews进行面板数据操作(有图有真相)

《使用Eviews进行面板数据操作(有图有真相)》由会员分享，可在线阅读，更多相关《使用Eviews进行面板数据操作(有图有真相)（44页珍藏版）》请在装配图网上搜索。

1、Eviews面板数据上机操作2011年4月26日、建立两种eviewsl作文件(1) 建立混合数据型工作文件1 建立一个年度工作文件(1996-2002) . (file:) filenewworkfileobjectnew objectpoolOKWorkfile CreateWorkfile structure typeD-ated - regular frequency Irregular Dated and Panel workfiles may be made from Un structured workfiles by later specifying date and/or o

2、ther iden tifier series Date specification! OK J I CancelNames (optional)WF；Page:在弹出的对话框中输入截面个体的名称缩写一一点击sheet,输入变量名 （注意：在面板数据中，变量名的后面要加上“？ ”，表示不同个体）。r13k在弹出的右图对话框中点击Edit键 输入数据，点击order键可以在“按 个体进行排序”和“按时间进行排 序”之间进行切换。生成新的序列打开原始数据（在变量列表中双击poolOl即可打开），点击工具栏倒数第二个的“poolGenr”键，或者选择quickgenerate series,在弹出的

3、对话框中输入要生成的公式，如:cp?=consume?/p?, ip?=income?/p? Pool: POOLOl Workfile: PANEL DATA:UntitledrView Proc Object Print Name Freeze ：Edit4-/- Order+/- Smpl4-/- Format Title Estimate Define PoolGenr SampleobsCONSUME? |INCOME?P?BJ-19965729.5207332.010111.60004ZZZ)VBJ-19976531.8107813.160105.3000BJ-19986970.8

4、308471.980102.4000BJ-19997498.4809182.760100.6000BJ-20008493.49010349.69103.5000BJ-20018922.72011577.78103.1000BJ-200210284.6012463.9298.20000TJ-19964679.6105967710109.0000TJ-19975204.1506608.390103.1000TJ-19985471.0107110.54099.50000TJ-19995851.5307649.83098.90000TJ-20006121.04081 40.50099.60000TJ-

5、20016987.2208958700101.2000TJ-20027191.9609337.56099.60000HB-19963424.3504442.810107.1000HB-19974003.7104958.670103.5000HB-19983834.4305084.64098.40000HB-1999dn?6 mririmnqa wnnnHR-9nnn4L卜(2) 建立面板数据型工作文件1 建立一个年度工作文件(1996-2002) . (file:)file new workfile 在弹出的对话框的左边选择Balance panel 并输入起始时间，在截面数据选项中输入T5”

6、 一一OKobjectnew objectpoolOK(2) objectnew objectquickempty group,后面步骤 的和截面数据的操作一样。Workfile CreateObject View ProcQuickOQtions Win dow HelpBalanced P曰 nelIrregul-ar Dated and Panel workfiles may be made from Unstructured workfiles by later specifying date and/or other identifier series Panel specifica

7、tionNumber of cross sections:15Workfile:Names (opti onS)WF；Page:Range: 1996Sample: 1996 ismebscrossid daieid resid ser01 ser02Sample.Generate Series.Show .Graph Empty Group (Edit Series)Series Statistics Group Statistics Estimate Equati on. Estimate VAR.注意：由于无法输入个体名称, 容易产生混乱，因此不建议使用二、画个体符号不同的散点图方法一:

8、在变量截面中按住Ctrl键，依次选择每个个体的因变量和自变量，点击右键, 选择openas group,在打开的数据窗口中选择viewgraphconsumehbOpenconsumejs consume consumeln& cpnmg Kcpsd2S cpshcpsx Scptj & cpzj0pln0pnmg EpoolOI0psdGroup: UNTITLED Workfile: PAPas Groupcon sume nmgconsumesh consumes0 income? 0 incomet Hincomef 0 incomet 0 incomer 0incomej Hinco

9、mej 0incomej;CopyPastePaste Special.Manage Links & Formulae.Fetch from DB.Store to DB.Object copy as Equation.as Factor.as VAR. asSystem.as Multiple seriesE3両5 Print Item 険甸 Group Members Spreadsheet Dated Data TableGraph.,Descriptive StatsCovarianize Analysis.N-Way Tabulation.Tests of Equality.Prin

10、cipal Comp on ents.IPAH .06251 i.40247 .70470 .78528 .56753 .40597 .93333ah切 Bhbhlj卩.)s p p p p p p p c c c c c c c slslslsls曲Is0 incomel0 incomer0 incomes Hincomesn HincomesxRename.Delete0incometj0incomezjH I S X I - Z /- - -C w- p p P在弹出的对话框中选择scatter在multiple中选择single graph-XY pairsOK便可得如下的散点图o (

11、CPAH.IPAH) i (CPFJJPFJ) x (CPHU.IFHLJ) (CPJSJPJS) x (CPLN.IPLN)(CPSD.IPSD)(cpsx.ipsxiI- (CPZJJPZ (CPBJ.IP0J) (CPHBJPHB) + (CPJL.IPJL) (CPJX.IPJX) a (CPNMG.IPNMG)(CPSH.IPSH) (CPTJJPTJ)方法二（个人认为该方法较麻烦）：新建立一个截面数据文件夹filenewworkfile,在弹出的对话框左侧选择unstructured/undated,右边输入 1 05OKobjectnewobjectseries,在右边变量名的对

12、话框中输入变量名cp_l 996 （这里一定要输入，否则不会出现该变量）一一0K双击变量名cp_1996,在弹出的 对话框的第1 -1 5个空格中输入数据。Workfile CreateWorkfib structure typeUnstructured / Undated D-ata r-angeObservations: 105Irregular Dated and Panel workfiles may be made from Un structured workfiles by later specifying date and/or other identifier series

13、CancelView Proc Object Properties113282.46625133.97834011.77543197.33952904.68762833.32173712.26082714.12493237.275102572.342113440.684126193.333132813.336144293.220155342.23416NA按照以上方法依次建立变量cp_1997、cp_1998，并输入数据，在输入 数据时cp_l 997要在第1 6-30个空格中输入，cp_l 998要在第31-45个空 格中输入，依次类推。然后再建立变量cp_iah、cpbj、cp.ifj方法

14、同上建立一个以所有的因变量值()为横坐标的变量ipSeries: CP IBJ Workfile:View Proc Object Properties PrintSeries: CP_1997 % Series: CP_20Q2 VVorkfile: UNTIView ProcObject Properties Print NamefView ProcObject ； Properties Print Name115NA89NA163646.1 5090NA176203.048914784.364184853.441921 0473.1 2193868.31 9936665.00520307

15、7.9899451 20.485213286.432954493.535224457.788964998.8742331 36.873976091.331243608.060984544.775252901.722995402.063263930.5741004850.1 80276634.1 831015635.770283131.62910210411.94295047.6721034787.561306002.0821047220.84331hJQ1058792.21 0qq6NA7NA85133.97896203.048106807.451117453 757128206.271138

16、654.4331410473.1215NA16NAView Proc | Object Properties90NA91NA924293.220935047.672945498.503955916.613966145.622976904.368987220.84399NA Series: CPJTJ Wori 建立以所有的因变量值()为横坐标的变量ip和ip_t,其中ip是按个体排序的 1 05个ip值，ip_t是按时间排序的1 05个ip值。 按Ctrl键，分别按顺序点击ip_t,cp_l 996,cp_l 997,cp_l 998一一右键 open as a group(或者show),此

17、时打开的数据为阶箱型，其中第一列为全 部105个ip值,后面分别为1996年、1997年.的cp值。1006038.9221017668.03610213183.881036335.7321049375.06010511822.001S? EViews Group: UNTITLED Workfile: SANDIANTUUntitled口 File Edit Object View Proc Quick Options Window HelpView | Proc| Object Print | Name | Freeze | | Def suit Sort | Transpose | Ed

18、it-t-/-1 Smpl+/-1 InsCobsIP TCP 1996CP 1997CP 1996CP 1999133431.5942813.336NANANA145474.9634293.220NANANA156446.6155342.234NANANA164540.247NA3646.150NAMA177419.905NA6203.048NANA186040.944NA4853.441NANA194790.986NA3868.319NANA203918.314NA3077.989NANA214041.061NA3286.432NANA225668.830NA4457788NAMA2339

19、91.490NA3136.873NANA244382.250NA3608.060NANA253774.804NA2901.722NANA265049.407NA3930.574NANA278209.037NA6634.183NANA283869.952NA3131.629NANA296409.690NA5047.672NANA307158.288NA6Q02.082NANA314770.470NANA3777.410NA328273.418NANA6807.451MA336505.145NANA5197.041NA345167.317NANA3896.778NA354251.494NANA32

20、89.990NA364240.565NANA3477.560NA374在打开的数据组中点击viewgraphscattersimple scatter 便可得到不同时间的散点图。同理，按Ctrl键，分别选择ip_i, ip_ah,l p_bj, ip_hb便可得到不同个体 的散点图。11000-10000-9000-8000-7000-6000-5000-4000-3000-6 7 89 9 99 9 9 1 -1 di _ _ _ ppp c c c9 0 0 12 2 p p P c c Co CPAH+ CP.JJL CP.JSD CP IBJ CPUSq CPJSH CP IFJo C

21、PIJXO CPISX* CPHB CPILN CP.JTJ:CPHLJA CPINMG0 CP.IZJ20002000 4000 6000 8000 10000 12000 14000IPJ由貓讒嚨豔洛泊三、配回归直线（file： 7panelO5食品支出）按住Ctrl,选择要画回归线的两个变量，在本例中按住Ctrl,选择fl和il openas group在打开的数据对话框中选择viewgraph在graph type选项中选择scatter在对画框右边点击option选项Group: UNTITLED Workfile: UNTGroup MembersSpreadsheetDated

22、Data TableMi60.950062.6800i83.4400i95.3000Graph.Descriptive Stats Covariance Analysis. N-Way Tabulation. Tests of Equality. Principal Components. i30.7500i33.8500i40.7700iSO.OOOOi67.5500i76.6400Correlogram (1).Cross Correlation (2) Unit Root Test.Cointegration Test.Qranger Causality.i82.2700834500；9

23、5.000004.1 60004.30001 7.7000Label29.840033.3100Option选项对话框中有多种选择（具体图形在下页）:a. x选inverse表示倒数模型b. x选log表示对数模型c. x、y都选log表示双对数模型d. x选power 2表示是只有平方项的模型e. x选polynomial 2表示既有平方项又有一次方项的模型倒数模型20.000-f双对数模型有一次方的二次项模型F1对数模型从这几个图形来看，双对数模型和 二次多项式模型拟合的较好，但是 图形越来越分散，说明存在异方差, 可以用下一页的方法来克服异方差。无一次方的二次项樓型(3) 克服二次多项式

24、模型的异方差：按Ctrl选择fl ,il show在对话框中输Alog(fl) log(il)打开数据窗 口viewgraphscatteroption在x transformation中选择polynomialOK,便可得无异方差的散点图。obsLOG(F1)LOG )15.6266856.32963225.7143606.33271135.6617106.36894245.5691086.38906555.712U76.44691065.7780246.45181275.7567546.462671|vigw|Pro.cobject Piint |rtam4|Frezg| | Defaul

25、t |10.09.5-9.0-8.5-8.0-6.5-7.08.57.58.06.06.56.05.50017.0-LOG(F1)9.0四、画拟合曲线点击之前所做的估计结果的名称（这里为pool02,每次做出结果后最好保存, 便于以后随时打开），在打开的结果窗口中点击procmake model在弹出的对话框中点击solve在弹出的Model solution对话框的左边中 间部分选择Static solution此时原始的变量窗口会多岀cpah_0 cpbj_O等，这是个体变量的拟合值。tlPool: P00L02 Workfile: 5PANEL02PaneView?rdq| Object

26、 1 Print 1 NameFreezeEstimateDefiModel: UNTITLED Workfile: 5PANEL02PanMigyv|Proc|obj&:t| Print|N3mg|FreeN&| Solw | EqMtions|SaveDel：3ils4/-| ShowI H ICro Tot?Li礙EstimateMMMake ResidualsMake Group.Make Period Stats series.1 Make ModelMake System.Update Coefs from PoolGenerate Pool series. Delete Pool

27、 series.R-sAdjiS.EP马Store Pool series (DB).Fetch Pool series (DB).Import Pool data (ASCILXLS.WK?).Export Pool datJ (ASCII,XLS,WK?).I Equati on s: 15E POOL02Eq1.DyTiAmic solution Static solutioni,r Fi t static - no eq inteStructural (ignore AJobsobs0 consumetj 0 consumezj EElcp0 cpah0 cpah_O0 cpbjS c

28、pbj_O0 cpflS cpij_O0 cphbS cphb_OS cphlj0 cphljJD3ISDIK K n n.w.wd.d.c c c c按住Ctrl,选择cpah和cpah.Oopen as a group在弹出的数据对话框中选择viewgraphline便可得到由原始值和拟合值曲线构成的图。如果想在图形中出现散点，可以双击图形, 在弹出的对话框中选择line&symbol,便可 出现带有散点图的图形。Line attributes五、面板数据的估计和检验(file :5panelO2)1.面板数据的估计 建立面板数据文件夹，具体步骤参考第一部分 打开数据窗口，点击estima

29、te,弹出pool estimate对话框，如下页所示。V填入因变量选择None表示混合 模型，选择fixed表 示个体固定效应模型, 选择random表示个 体随机效应模型选择None表示混合 模型，选择fixeH表 示时间固定效应標型, 选择random表示时 间随机效应模型GLS权重，通过加权可以克服异方差变参模 型V下面为个体固定效应的结果。点击viewrepresentation可以显示具体的回归方程式。Pool: POOL02 Workfile: 5PAN迈鐵 I Prod Object | Print Namere 述 | _E Cross Section Identifiers

30、RepresentationsS? EViews - Pool: POOL02 Workfile: 5PANEL02Panel02口 File Edit Object View 臼cc Quick Options WindowView | Pfoc | Object|Estirate | Define PoolGenr | Sheet |VariableCoefficie ntStd Errort-StatisticC533440369.347727.692266IP?0.6947360.01086363.95600Fixed Effects (Cross)AH-C-3977133BJ-C54

31、6.0966FJ-C-45.58037HB-C-156.7460HU-C-174.2270JL-C19.93251JS-C-34.68271Estimation OutputResidualsCoef Covariance MatrixConfidence Ellipse.Wald Coefficient Tests.View | Proc object | Print Name Freeze | Estimate | Define | PoolGenr | Sheet | buDstituted uoetticierrts:CPAH = -39.77133443 + 533.440323 +

32、 0.6947364855*IPAH:PBJ = 546.0965643 + 533.440323 十 0.6947364855*IPBJCPFJ = -45.58037043 + 533.440323 + 0.6947364855*IPFJCPHB = -156.7459542 十 533.440323 + 0.6947364855*IPHBCPHLJ 二-174.2269943 十 533.440323 十 0. 6947364855*IPHLJCPJL = 19.93250903 + 533.440323 + 0.6947364855*IPJL2.面板数据的检验Hausman检验（要在随

33、机效应结果窗口中进行）对数据进行随机效应模型估计，在估计结果窗口点击viewFixed/RandomEffects testingCorrelated Random Effect-Hausman Test （6.0以上白勺版本才可以）Pool: POOL02 Workfile: 5PANE：Pool: POOL02 Workfile: 5PANEL02:Panel02計磴yy Proc Object Print Name Freeze Estimate Define PoolGenr SheetCross Section IdentifiersC345.1795IP?0.724569Rando

34、m Effects (Cross)AH-C-2.553433BJ-C367.0439FJ-C-54.24006HB-C-104.8367HU-C-101.7680JL-C54.90671JS-C-32.27868JX-C-223.9519LN-C112.1152NMG-C-133.1377SD-C-100.8713SH-C126.1820SX-C-22.79189TJ-C10.08794ZJ-C106.0939站如Ptxobj&d: print加皂总対 EsQ芒iiV七PU巧【Hi亡仃iFtearesentationsEstimation OutputResidualsCoef Covaria

35、ncE MatrixCoeffi匚ient TestsFixed/Randorn Effects TestingSpreadsheet (stacked data). Descriptive Statistics .Unit Root Test.Cointegration Test.Label)5lent variancesStd. Error 1Statistic Prob.Redundant Fixed Effects - Likelihood RatioCorrelated Random Effects Hausman Test注意：只有在随机效应估计窗口中才能 进行Hausman检验，

36、只有在固定效应估 计窗口中才能进行似然比检验Pool: P00L02 Workfile: 5PANEL02:Panel02Yi 胡Object|Piint|NameFr88ze Estimatepefin&poolGend品estTest SummaryChi-Sq. StatisticChi-Sq. d.f.Prob.Cross-section random14.78756310.0001Correlated Random Effects HausmanTostPool: POOL02Test cross-section random effectsHausman检验的原假设是个体效 应与回

37、归变量无关，应建立随机效 应模型，因此当Hausman值较大， 其对应的P值远小于0.05时，拒绝 原假设，应建立个体固定效应模型。似然比检验方法同上，只是要在个体固定效应模型的输出结果下进行检验*liPool: POOL02 Workfile: 5PANEL02:Panel02F曰View |PrQcj ObjectPrint|Nane| 屁 e 乙 eEstimate Define jP OQlGenr ： Sh eetEffects TestStatisticd.f.Prob.Cross-section F7.151832(14.89)0.0000Cross-section Chi-sq

38、uare79.146400140.0000Redundant Fixed Effects TestsPool: POOL02Test cross-sectionfbcod effects该检验的原假设是勺相等，应建 立混合效应模型，当F值较大，P 值远小于0.05时，拒绝原假设， 应建立个体固定效应模型。两个检验的结果都是要建立个体固定效应模型, 因此这组数据应建立个体固定效应模型。 其他功能点击 viewResidualTable,Graphs, Covarianee Matrix, Correlation Matrix可以分别得到 按个体计算的残差序列表、残差序 列图、残差序列的方差协方差

39、矩阵、 残差序列的相关系数矩阵。在估计结果中点击procMakeModel可以出现估计结果的联立方 程形式，进一步点击Solve键可以 在弹出的对话框中进行动态和静态 预测。DeFSar Incl CroTote Poof: P00L02 Workfile: 5PANEL02Panel02迅議| Proc|Object| Print) Name|Freeze; Estimate Define|PoolGenr |SheetCross Section IdentifiersRearese ntationsEstimation OutputaResiduals 1TableCoef C ovari

40、ance Matrix Confiden 匚 e Ellipse. Wald Coefficient Tests.GraphsCovariance MatrixCorrelation MatrixSpreadsheet (stacked data) .Descriptive Statistics .Unit root test.ValueStd. Err.Label514.9142 81.59665Restrictions are linear in coefficients.Pool: P00L02 Workfile: 5PANEL02Panel0View ?rdg| Object Prin

41、t Name Freeze Estimate DefineEstimateMake ResidualsMake Group.Make Period Stats series.Make ModelMake System.Update Coefs from PoolGenerate Pool series.Delete Pool series.Store Pool series (DB).Fetch Pool series (DB).Import Pool data (ASCII,XLSUWK?).Export Pool data (ASCII,XLS,WK?).五、面板数据的单位根检验在估计结果

42、或原始的面包数据窗口中点击viewunit root testPool: POOL02 Workfile: 5PANUnit Root TestI Proc! Object Print Name Freeze 巳Cross Section IdentifiersRepresentationsEstimation OutputResidualsCoef Covarinup Matrix Confidence Ellipse.Wald Coefficient Tests.Spreadsheet (stackeddata). Descriptive Statistics.Unit root test

43、.Lbel这里的检验方式为综 合检验，包括LLC、 Breitung、ADF等， 其中Hadri检验的原 假设是没有单位根， 其余燮竺勺原假设都这里默认为 Schwarz 检 验，因为在 小样本情况 TSchwarz 检验效果最 好。这里也可以选择单位根一阶滞 后或二阶滞后MethodStatisticProb.*CrosssectionsObsNull: Unit root (assumes common unit rootLevin. Lin & Chu t*9.697781.00001584Breitung t-stat-4.894170.000015Null: Unit root (a

44、ssumes individual unit rooprocess)Im. Pesaran and Shin W-stat6.463801.00001584ADF - Fisher Chi-square3.049971.00001584PP - Fisher Chi-square6.243351.00001590Null: No unit root (assumes common unitDot proce?Hadri Z-stat7.5S3830.000015105除 了 Breitung 检验外，其他 检验都说明有 单位根，按照 少数服从多数 原则，这里cp 具有单位根。Pool: POO

45、LO2 Workfile: 5PANEL02Panel02ViEw | Proc | Object | Print | N占mm | Freeze | EstimEtm | Define | PoolGmrr | Shmet |Pool Unit Root Test on CP?Pool unit root test: SummaryDate: 04/26/11 Time; 21:40Sample: 1996 2002Series: CPAH. CPBJ. CPFJ. CPHB, CPHLJ. CPJL. CPJS. CPJX.CPLN. CPNMG. CPSD. CPSH. CPSX. CPTJ. CPZJ Exogenous variable3： InTZidu召I effects Automatic seleclion of maximum lagsAutomatic selection of lags based on SIC: 0 to 1N ewe y-We st bandwidth selection using Bartlett kernel