Eviews實(shí)際案例：序列相關(guān)性

1、 碎客天涯之計(jì)童經(jīng)濟(jì)學(xué)第二步:1將估計(jì)的p=0.6278代入差分模型，得到廣義模型，進(jìn)行eviews運(yùn)行2輸入公式為:log(y)-0.6278*log(y(-l)clog(x)-0.6278*log(x(-l)3運(yùn)行結(jié)果：DependentVariable:LOG(Y-0.6278*LOG(Y(-1)Method:LeastSquaresDate:05/14/12Time:21:23Sample(adjusted);19812000Includedobservations:20afteradjustmentsCoefficientStd.Errort-StatisticProb.C0.425

2、6570.1301073.2715940.0042LOG(X)-0.6278*LOG(X(-1)0.9017320.03809523.670850.0000R-squared0.968875Meandependentvar3.481639AdjustedR-squared0.967146SD.dependentv日0.397956SEofegression0.072133Akaikeinfocriterion-2.325982Sumsquaredresid0.093656Schwarzcriterion-2.226408Loglikelihood25.25982Hannan-Quinncrit

3、er.-2.306544F-statistic560.3093Durbin-Watsonstat1.393141Prob(F-statistic)0.000000最小二乘估計(jì)結(jié)果為:-0.6278111=0.4257+0.9017(lnX,-0.6278111口)(3.272)(23.671)D.W.=1393由于D.WR.393,在5%的顯著性水平下,樣本容量為19的D.W.檢驗(yàn)的臨界值為上卜限為dL=1.18,*=1.40,檢驗(yàn)值落在(,*)上，故不能確定是否冇在一階序列相關(guān)。根據(jù)LM法進(jìn)行檢驗(yàn)。在eviews輸出結(jié)果窗II選擇ViewResidualTestsSeriesCorrelat

4、ionLMTest，在隨后出現(xiàn)的對話框中填入滯后期數(shù)“2”，得到：Breusch-GodfreySerialCorrelationLMTest:F-statistic1.131672Prob.F(1,17)0.3023ObsR-squared1.248282Prob.Chi-Squared)0.2639檢驗(yàn)統(tǒng)計(jì)量值為1.248,查詢Z2分布，r005(l)=3.841-=2.59,檢驗(yàn)值小于臨界值,因而不能拒絕原假設(shè)，認(rèn)為模型不存在一階序列相關(guān)。八04257因此估計(jì)原模型為：lnX二一+0.9017InX,1-0.6278即：111=1.156+0.9017111(二)廣義最小二乘法在Evie

5、ws中,選擇uQuickEstimateEquationn，輸入log(y)clog(x)AR”AR(1)表示隨機(jī)干擾項(xiàng)是一階自回歸形式的序列相關(guān)。運(yùn)行結(jié)果如下:DependentVariable:LOG(Y)Method:LeastSquaresDate:05/14/12Time:22:24Sample(adjusted):19812000Ineludedobservations:20afteradjustmentsConvergenceachievedafter20iterationsCoefficientStd.Errort-StatisticProb.C1.1642330.413568

6、2.8150970.0119LOG(X)0.8994470.04444820.235880.0000AR0.6451280.1517384.2515930.0005R-squared0.995321Meandependentvar9.102328AdjustedR-squared0.994771S.D.dependentvar1.035997S.E.ofegession0.074916Akaikeinfocriterion-2.207425Sumsquaredresid0.095410Schwarzcriterion-2.058065Loglikelihood25.07425Hannan-Qu

7、innenter.-2.178268statistic1808.248Durbin-Watsonstat1.407464Prob(F-statistic)0.000000InvertedARRoots.65根據(jù)LM法進(jìn)行檢驗(yàn)。在eviews輸出結(jié)果窗II選擇uViewResidualTestsSeriesCorrelationLMTest，在隨后出現(xiàn)的對話框中填入滯后期數(shù)1,得到：Breusch-GodfreySerialCorrelationLMTestF-statistic1.707900Prob.F(1,16)0.2097ObsrR-squared1.928969Prob.Chi-Squ

8、ared)0.1649同上，可以判定模型隨機(jī)干擾項(xiàng)中不再存在一階序列相關(guān)，因而廣義最小二乘估計(jì)結(jié)果為：ln=1.164+0.8994InX,+0.6451AR(1)三、遙用差分形芬乂；=兀-鼻與作為新數(shù)據(jù)，估計(jì)模型Y：=%+,該模型是否存在序列相關(guān)選擇“QuickEstimateEquationn，在出現(xiàn)的對話框中輸入“D(Y)CD(X)”，其中D(Y)與D(X)分別表示對序列Y與X取差分。得到運(yùn)行結(jié)果如下；DependentVariable:D(Y)Method:LeastSquaresDate:05/14/12Time:22:40Sample(adjusted):19812000Inclu

9、dedobservations:20afteradjustmentsCoefficientStd.Errort-StatisticProb.C411.0327384.80721.0681520.2996D(X)0.9180280.1741135.2725910.0001R-squared0.606989Meandependsntua1880.190AdjustedR-squared0.585155S.D.dependentvar1842.763S.E.ofregression1186.896Akaikeinfocriterion17.09071Sumsquaredresid25356983Schwarzcriterion17.19028Loglikelihood-168.9071Hannan-Qu