1、C-10.3934186.05105-0.1207820.9047GDP0.0710320.0074069.5912490.0000R-squared0.760315Mean dependent var621.0548Adjusted R-squared0.752050S.D. dependent var619.5803S.E. of regression308.5175Akaike info criterion14.36377Sum squared resid2760308.Schwarz criterion14.45629Log likelihood-220.6385F-statistic
2、91.99205Durbin-Watson stat1.570581Prob(F-statistic)0.000000回归方程为:Y = -10.39340931 + 0.07103165248*GDP 对回归方程做检验:斜率项t值9.59大于t在5%显著水平下的检验值2.045,拒绝零假设;截距项t值0.121小于2.045,接受零假设。可决系数0.76,拟合较好,方程F检验值91.99通过F检验。下面进行预测:拓展工作空间:打开work file窗口,单击 ProcStructure,将End date的数据3132;确定预测值的起止日期:打开work file窗口,点击QuickSamp
3、le,填入“1 32”。打开GDP数据表,在GDP的最下方填,按回车键。在出现的Equation界面,点击Forecast出现相应界面如下:双击YF,得到y32=593.3756,预测完毕。实验二:回归模型的建立与检验(数据来源于李子奈版课后习题P105.11)运行Eviews,依次单击filenewwork fileunstructedobservation 10。命令栏中输入“data y x1 x2”,打开“y x1 x2”表,接下来将数据输入其中。命令栏中输入“ls y c x1 x2”回车,即得到回归结果: 10:17 1 10 10626.509340.1301015.61195X
4、1-9.7905703.197843-3.0616170.0183X20.0286180.0058384.9020300.00170.902218670.33000.87428149.0450417.389858.7929752116.8478.883751-40.9648832.294081.6508040.000292估计方程:依次单击viewrepresentations,得到回归方程为: Y = 626.5092847 - 9.790570097*X1 + 0.02861815879*X2,参数估计完毕。直接查看结果计算得到随机干扰项的方差值为2116.847/(10-2-1)=309
5、.55,可决系数为0.902,修正后的可决系数为0.874。F=32.2945%显著水平下的F值4.74,即方程通过F检验;两个参数的t检验值均通过了5%显著水平下的t检验值2.365。打开work file窗口,单击 ProcStructure,将End date的数据1011;打开work file窗口,点击QuickSample,填入“1 11”。在x1的最下方填入35,在x2的最下方填入20000,按回车键。双击YF,得到y11=856.2025,预测完毕。实验三:异方差、自相关、多重共线性的检验1.异方差检验(数据来源于李子奈版课后习题P154.8)运行Eviews,依次单击file
6、newwork fileunstructedobservation 20。命令栏中输入“data y x”,打开“y x”表,接下来将数据输入其中。开始进行LS回归,命令栏中输入“ls y c x”回车,即得到回归结果如下: 1 20 20272.3635159.67731.7057130.1053X0.7551250.02331632.386900.9831295199.5150.9821921625.275216.890013.69130846743.013.79087-134.91301048.9122.087986Y = 272.3635389 + 0.7551249391*X开始检验
7、异方差图示法:在工作文件窗口按Genr,在主窗口键入命令e2=resid2,依次单击QuickGraphScatter可得散点图:显然,散点不在一条水平直线上,即说明存在异方差性。White检验法:依次单击ViewResidual TestsWhite Heteroskedasticity因为本题为一元函数,故无交叉乘积项,选no cross terms。经估计出现white检验结果,如下图:White Heteroskedasticity Test:14.63595Probability0.000201Obs*R-squared12.652130.001789Test Equation: R
8、ESID2 11:16-180998.9103318.2-1.7518580.097849.4284628.939291.7080060.1058X2-0.0021150.001847-1.1447420.26820.63260642337.150.58938445279.6729014.9223.526491.43E+1023.67585-232.26492.081758所以拒绝原假设,表明模型存在异方差。 Goldfeld-Quanadt检验法:在命令栏中直接输入:sort x,得到按照升序排列的x。开始取样本,依次单击quicksample,填入“1 8”,回归模型ls y c x;得到
9、如下结果:26 1 8 81277.1611540.6040.8290000.43880.5541260.3114321.7792870.12550.3453974016.8140.236296166.1712145.217213.00666126528.313.02652-50.026633.1658613.0045320.125501继续取样本,依次单击quicksample,填入“13 20”,回归模型ls y c x;28 13 20212.2118530.88920.3997290.70320.7618930.06034812.625050.9637236760.4770.95767
10、61556.814320.279014.58858615472.014.60844-56.35432159.39191.7229600.000015计算F统计量:F=RSS2/RSS1=615472.0/126528.3=4.864;F=4.864 F0.05(6,6)=4.28,拒绝原假设,表明模型确实存在异方差性。异方差的修正:在对原模型进行OLS后,单击QuickGenerate Series,在弹出的对话框内输w1=1/e,w2=1/e2。再选择QuickEstimate Equation,在弹出的对话框中选择Options按钮,在出现的画面中,选中Weight Ls/TLS复选框,在
11、 Weight内分别输入“w1”,“w2”,分别得下图:33Weighting series: W1415.6603116.97913.5532880.00230.7290260.02242932.50349Weighted Statistics0.9998954471.6060.9998897313.16077.0483111.62138106856.011.72096-114.21381056.4772.367808Unweighted Statistics0.9816640.980645226.1101920263.91.88695934 W2117.0597134.71860.8689
12、200.39630.7869760.02605830.200730.9999994207.51615774.2115.359928.3960394246.6888.495613-81.96039912.08392.1136590.9802810.979185234.4844989692.91.836717经估计发现用w2=1/e2作为合适的权。再检验:单击QuickGenerate Series,分别输入x1=x*w2,y1=y*w2,按住ctrl,依次点击x1,y1,右键选择Open as group,依次单击QuickGraph可得下图:由该图可知,加权后X和Y的散点图在同一直线上,所以是同方差性。2.自相关检验(数据来源于李子奈版课后习题P155.9)运行Eviews,依次单击filenewwork fileAnnualstrat1980 end2007。开始进行LS回归,命令栏中输入“ls log(y) c log(x)”回车,即得到回归结果如下: LOG(Y)49 1980 2007 281.5884780.13422011.83492LOG(X)0.8544150.01421960.090580.9928519.5522560.9925761.3039480.112351Akaike inf