课程代码：学时/学分： 成绩：计量经济学课程案例分析案例主题：农业生产的多种投入要素对农业总产量及总产值影响的实证分析 任课老师：组 长： 组 员： 计量经济学小组实验报告目 录计量经济学小组实验报告1u截面数据31.单方程模型3一、模型设定3二、参数估计和模型检验5三、多重共线性检验6四、异方差性检验12五、自相关性检验142.联立方程模型14一、实验目的14二、本实验选用模型及数据说明15三、单方程估计实验过程16四、系统估计方法实验过程22五、实验小结25u时间序列数据25一、数据说明25二、滞后模型26三、检验序列相关性33四、单位根检验34五、构造ARMA模型36六、EG两步法协整检验38七、格兰杰因果检验39八、误差修正模型40u 截面数据1. 单方程模型一、 模型设定根据经济学尝试，农业产量受自然环境和人为因素的影响。在不同气候条件下，不同作物和不同的灌溉、化肥等都会对该地区的农业总产量产生影响。在本文中，考察农业总产量的的主要因素包括，播种面积、灌溉面积和化肥使用量。因为天气因素太不稳定，有很大的随机性，同时也不容易量化，所以这里没有考虑。综合上述因素和变量的可观测性，我们首先建立一个多元线性回归模型：其中：ZCL：农业总产量BZMJ：播种面积GGMJ：灌溉面积HUAF：化肥使用面积 是回归系数，是随机扰动项收集数据如下表表示：ZCLBZMJGGMJHUAF109.2000386.4000322.700015.70000147.2000544.5000354.300017.300002645.6008990.8004485.400273.4000850.20003672.3001104.30084.900001319.7005707.3002472.30079.300001440.7003964.8001482.800109.80001987.7004890.1001382.600114.10002688.0009989.2002090.400123.2000164.2000490.9000280.600020.300003174.6007777.4003900.000338.00001130.9003245.9001400.30090.300002799.1008733.1003228.700280.7000843.40002713.100942.4000117.40001690.5005534.7001897.500109.70004097.30011266.104836.100428.60004482.50013127.704766.000441.70002417.9007489.0002027.900245.30002937.7007931.7002676.300184.30001681.0005193.1001447.100195.10001568.6006288.1001519.600168.1000206.1000871.7000180.800027.000001053.5003555.900631.900072.600003107.5009571.5002533.000212.00001171.6004650.700659.800070.000001514.0005929.6001424.300120.0000102.7000230.9000154.40003.0000001014.1004331.9001314.100131.0000791.60003688.900982.300066.10000126.2000529.0000208.30007.200000282.10001007.600405.400024.60000822.60003404.1003138.10083.30000二、 参数估计和模型检验直接对模型进行OLS法估计，结果如下：Dependent Variable: ZCLMethod: Least SquaresDate: 09/02/08 Time: 20:10Sample: 1 31Included observations: 31VariableCoefficientStd. Errort-StatisticProb.C-83.4463175.57143-1.1042040.2792BZMJ0.2341980.0288108.1290520.0000GGMJ-0.0149690.074818-0.2000780.8429HUAF3.5967070.9554373.7644620.0008R-squared0.966425Mean dependent var1560.258Adjusted R-squared0.962694S.D. dependent var1205.894S.E. of regression232.9156Akaike info criterion13.85914Sum squared resid1464741.Schwarz criterion14.04417Log likelihood-210.8167Hannan-Quinn criter.13.91946F-statistic259.0533Durbin-Watson stat1.762135Prob(F-statistic)0.000000分析由上表我们看到解释变量的t值不显著，而可决系数F和R统计量显著，说明极有可能存在多重共线性。三、 多重共线性检验计算解释变量之间的简单关系系数，结果如下：ZCLBZMJHUAFZCL1.0000000.9709010.934915BZMJ0.9709011.0000000.890866HUAF0.9349150.8908661.000000由上表可以看出，各解释变量之间有严重的多重共线性。多重共线性修正用OLS法逐一求ZCL对各个解释变量的回归，发现ZCL对BZMJ的线性关系很好，拟合度最好，如下：VariableCoefficientStd. Errort-StatisticProb.C-132.511693.78046-1.4129980.1683BZMJ0.3370150.01543621.832320.0000R-squared0.942648Mean dependent var1560.258Adjusted R-squared0.940670S.D. dependent var1205.894S.E. of regression293.7275Akaike info criterion14.26552Sum squared resid2502000.Schwarz criterion14.35804Log likelihood-219.1156Hannan-Quinn criter.14.29568F-statistic476.6502Durbin-Watson stat1.858925Prob(F-statistic)0.000000但是，此时常数项C即原式中的的t值和p值都不显著，而且为严重的负相关不合常理，所以去掉常数项，因为它可能没有经济意义。去掉常数项后，BZMJ的各项检验值都有所提高，如下：VariableCoefficientStd. Errort-StatisticProb.BZMJ0.3189810.00882736.138380.0000R-squared0.938700Mean dependent var1560.258Adjusted R-squared0.938700S.D. dependent var1205.894S.E. of regression298.5663Akaike info criterion14.26759Sum squared resid2674255.Schwarz criterion14.31385Log likelihood-220.1476Hannan-Quinn criter.14.28267Durbin-Watson stat1.603563下面，进行逐步回归，将其余解释变量逐一代入：将GGMJ带入后：VariableCoefficientStd. Errort-StatisticProb.BZMJ0.2709120.0282569.5879460.0000GGMJ0.1382380.0774751.7843030.0848R-squared0.944764Mean dependent var1560.258Adjusted R-squared0.942859S.D. dependent var1205.894S.E. of regression288.2592Akaike info criterion14.22794Sum squared resid2409708.Schwarz criterion14.32045