計量經濟學-自相關性的檢驗及修正

1、經濟計量分析實驗報告、實驗項目自相關性的檢驗及修正、實驗日期2015.12.13三、實驗目的對于國內旅游總花費的有關影響因素建立多元線性回歸模型，對變量進行多重共線性的檢驗及修正后，對隨機誤差項進行異方差的檢驗和補救及自相關性的 檢驗和修正。四、實驗內容建立模型，對模型進行參數估計，對樣本回歸函數進行統(tǒng)計檢驗，以判定估 計的可靠程度，包括擬合優(yōu)度檢驗、方程總體線性的顯著性檢驗、變量的顯著性 檢驗，以及參數的置信區(qū)間估計。檢驗變量是否具有多重共線性并修正。檢驗是否存在異方差并補救。檢驗是否存在相關性并修正。五、實驗步驟1、建立模型。以國內旅游總花費Y作為被解釋變量，以年底總人口表示人口增長水平，

2、 以旅行社數量表示旅行社的發(fā)展情況，以城市公共交通運營數表示城市公共交通 運行狀況，以城鄉(xiāng)居民儲蓄存款年末增加值表示城鄉(xiāng)居民儲蓄存款增長水平。2、模型設定為： t :0- 1 X 箕_2 X 2t _ 3 X 3t 工罪 t其中：K國內旅游總花費（億元）又宜一年底總人口（萬人）X2t一旅行社數量（個）X3t一城市公共交通運營數（輛）x4t-城鄉(xiāng)居民儲蓄存款年末增加值（億元）3、對模型進行多重共線性檢驗。4、檢驗異方差是否存在并補救。5、檢驗自相關性是否存在并修正。六、實驗結果消除多重共線性及排除異方差性之后的回歸模型為：Y?=-301.8388+ 0.382963X2檢驗I、圖示法1、6，0散

3、點圖1,500 *1,000 _.500 . * * 電E0 .* *部-500 ,*-1,000 ,-1,500 -I.二11-2,000-1,00001,0002,000ET(-1)大部分落在第I,田象限，表明隨機誤差項存在正自相關。2、et折線圖RESIDn、解析法1、D-W$驗Dependent Variable: COSTMethod: Least SquaresDate: 12/13/15 Time: 20:35Sample (adjusted): 1994 2008Included observations: 15 after adjustmentsVariableCoeffic

4、ientStd. Errort-StatisticProb.C-301.8388394.3549-0.7653990.4577AGENT0.3829630.03223111.881750.0000R-squared0.915681Mean dependent var3875.880Adjusted R-squared0.909195S.D. dependent var2295.093S.E. of regression691.6017Akaike info criterion16.03946Sum squared resid6218068.Schwarz criterion16.13387Lo

5、g likelihood-118.2960Hannan-Quinn criter.16.03846F-statistic141.1760Durbin-Watson stat0.641734Prob(F-statistic)0.000000D-W=0.641734查表知：dL =1.08, du =1.36所以存在一階正自相關2、LM檢驗Breusch-Godfrey Serial Correlation LM Test:F-statistic4.492419Prob. F(2,11)0.0375Obs*R-squared6.743741Prob. Chi-Square(2)0.0343Test

6、 Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 12/13/15 Time: 20:43Sample: 1994 2008Included observations: 15Presample missing value lagged residuals set to zero.VariableCoefficientStd. Errort-StatisticProb.C-171.0092361.8256-0.4726290.6457AGENT0.0213550.0333880.6396130.5355RESID(-1)0

7、.8902450.3137952.8370260.0162RESID(-2)-0.1256190.426001-0.2948800.7736R-squared0.449583Mean dependent var-6.37E-13Adjusted R-squared0.299469S.D. dependent var666.4441S.E. of regression557.7986Akaike info criterion15.70905Sum squared resid3422532.Schwarz criterion15.89786Log likelihood-113.8179Hannan

8、-Quinn criter.15.70704F-statistic2.994946Durbin-Watson stat2.036592Prob(F-statistic)0.077146nR2=6.743741,查表得 殍（p）=5.99。p值=0.0343小于0.05,拒絕原假設,不存在高階自相關。修正（迭代法）Dependent Variable: COSTMethod: Least SquaresDate: 12/13/15 Time: 20:55Sample (adjusted): 1995 2008Included observations: 14 after adjustmentsC

9、onvergence achieved after 32 iterationsVariableCoefficientStd. Errort-StatisticProb.C-105.10801522.424-0.0690400.9462AGENT0.3966590.1023713.8747320.0026AR(1)0.7549890.2872182.6286300.0235R-squared0.945391Mean dependent var4079.621Adjusted R-squared0.935462S.D. dependent var2236.535S.E. of regression

10、568.1771Akaike info criterion15.71015Sum squared resid3551077.Schwarz criterion15.84709Log likelihood-106.9711Hannan-Quinn criter.15.69748F-statistic95.21571Durbin-Watson stat1.467537Prob(F-statistic)0.000000Inverted AR Roots.75D-W=1.467537無法查表，所以用LM檢驗Breusch-Godfrey Serial Correlation LM Test:F-sta

11、tistic1.019016Prob. F(1,10)0.3366Obs*R-squared1.294691Prob. Chi-Square(1)0.2552Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 12/13/15 Time: 21:01Sample: 1995 2008Included observations: 14Presample missing value lagged residuals set to zero.VariableCoefficientStd. Errort-Statistic

12、Prob.C11.127301521.1400.0073150.9943AGENT-0.0335630.107550-0.3120670.7614AR(1)-0.3234740.430157-0.7519900.4694RESID(-1)0.4854860.4809351.0094630.3366R-squared0.092478Mean dependent var7.45E-06Adjusted R-squared-0.179779S.D. dependent var522.6469S.E. of regression567.6866Akaike info criterion15.75597Sum squared resid3222681.Schwarz criterion15.93856Log likelihood-106.2918Hannan-Quinn criter.15.73907F-st