关于影响我国南方几省市农业总产值因素的实证分析.doc

关于影响我国南方几省市农业总产值因素的实证分析问题简述：本文通过对我国南方几省市（包括上海、江苏、浙江、安徽、福建、江西、山东、湖北、湖南、广东、广西、海南、重庆、四川、贵州、云南）在2004年度农业总产值、农业劳动力、有效灌溉面积、农用化肥施用量、农村居民家庭生产性固定资产以及农业机械拥有量的统计数据的收集和整理，建立我国南方地区产出线性计量经济模型，并对模型中是否存在违反古典假设的情况（包括“多重共线性”、“异方差性”和“自相关性”）进行了多种方式的检验分析。然后对症下药，针对模型中所存在的问题选用适当的方法进行修正。最后应用产业经济学和区域经济学的相关知识对修正后的模型进行分析，解释其实际的经济含义，并对其反映出来的现实问题提出几点看法和建议。一、 模型假设（古典假设）1. 符号约定：分别对应上海、江苏、浙江、安徽、福建、江西、山东、湖北、湖南、广东、广西、海南、重庆、四川、贵州、云南这16个省市在2004年度的农业总产值（单位：亿元）分别对应16个省市在2004年度的农业劳动力（单位：万人）分别对应16个省市在2004年度的有效灌溉面积（单位：万公顷）分别对应16个省市在2004年度的农用化肥施用量（单位：万吨）分别对应16个省市在2004年度的农村居民家庭生产性固定资产（单位：元）分别对应16个省市在2004年度的农业机械拥有量（单位：万千瓦）随机扰动项序列残差序列2. 解释变量是一组固定的值，即是非随机的。3. 解释变量无测量误差。4. 模型自身不存在设定误差。5. 零均值假定，即在给定的的条件下，的条件期望值为零，即6. 同方差假定，即对于每一个给定的，的条件方差都等于一个常数，即7. 无自相关性假定，即不存在自相关，或中个项预测值互不影响，即8. 随机扰动项与解释变量不相关，即9. 正态性假定，即假定服从均值为0，方差为的正态分布，表示为二、 统计数据收集整理的统计数据如下地区农业总产值（亿元）农业劳动力（万人）有效灌溉面积(万公顷）农用化肥施用量单位：(万吨)农村居民家庭生产性固定资产（元）农业机械总动力（单位：万千瓦） 上 海 98.271.74257.3115.871687.35112.61 江 苏 981.21230.293840.98334.673290.253029.10 浙 江 529.4872.961403.8090.383738.942039.66 安 徽 617.91860.573285.38281.283908.283544.66 福 建 466.8735.92939.95120.292958.61951.91 江 西 383.7971.261873.16110.982398.181220.52 山 东 1599.32264.624760.79432.654733.428336.70 湖 北 733.41110.712043.69270.322431.091661.75 湖 南 671.71997.672675.34188.332191.082664.45 广 东 851.71543.411315.93199.612166.261788.80 广 西 500.81541.021516.67183.692305.271696.30 海 南 152.7187.25177.2733.924289.36221.62 重 庆 270.1813.19649.6971.602036.17695.67 四 川 804.72413.992503.15208.393092.471891.06 贵 州 275.51322.10682.7174.922714.07761.99 云 南 433.91690.221457.00129.224056.211542.91拟合使用的方法：最小距离法（用Eviews软件实现）三、 模型建立初始状态下的农业产出线性模型：用Eviews软件得到的分析报告如下Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:07Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C242.6481186.09921.3038640.2215X10.0581570.0965900.6021040.5605X20.0797500.0793021.0056490.3383X30.0072420.3007750.0240760.9813X4-0.0417030.058281-0.7155370.4906X50.1219870.0538902.2636470.0471R-squared0.860155 Mean dependent var585.6875Adjusted R-squared0.790233 S.D. dependent var368.9720S.E. of regression168.9905 Akaike info criterion13.37756Sum squared resid285578.0 Schwarz criterion13.66728Log likelihood-101.0205 F-statistic12.30156Durbin-Watson stat3.062212 Prob(F-statistic)0.000521四、 模型检验与修正1. 对解释变量之间多重共线性的检验：（1）简单相关系数矩阵法：用Eviews得到的协方差矩阵如下 X1X2X3X4X5X1 1.000000 0.715554 0.481862 0.258392 0.648671X2 0.715554 1.000000 0.422811 0.426337 0.885707X3 0.481862 0.422811 1.000000 0.178946 0.430369X4 0.258392 0.426337 0.178946 1.000000 0.549919X5 0.648671 0.885707 0.430369 0.549919 1.000000大致上可以判断出，之间可能存在共线形。（2）变量显著性和方程显著性的综合判断：在显著性水平，样本容量的条件下，t统计量的临界值为。由此可见，除了以外其余变量均是不显著的，所以变量之间存在多重共线性。（3）对多重共线性进行修正：变换模型形式：回归分析报告如下：Dependent Variable: LNYMethod: Least SquaresDate: 05/17/05 Time: 19:45Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C2.6849722.2942341.1703130.2690LNX1-0.0895820.196497-0.4558950.6582LNX20.0380250.2620960.1450790.8875LNX30.0596810.1351520.4415830.6682LNX4-0.1558380.284183-0.5483710.5955LNX50.6626690.3028392.1881910.0535R-squared0.911189 Mean dependent var6.170439Adjusted R-squared0.866784 S.D. dependent var0.704874S.E. of regression0.257270 Akaike info criterion0.402616Sum squared resid0.661879 Schwarz criterion0.692336Log likelihood2.779075 F-statistic20.51983Durbin-Watson stat3.049254 Prob(F-statistic)0.000058可见效果并不明显，因此将采用逐步回归法进行修正：Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:48Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C95.40032150.91780.6321340.5375X10.3803090.1043523.6444790.0027R-squared0.486845 Mean dependent var585.6875Adjusted R-squared0.450192 S.D. dependent var368.9720S.E. of regression273.5893 Akaike info criterion14.17760Sum squared resid1047916. Schwarz criterion14.27418Log likelihood-111.4208 F-statistic13.28223Durbin-Watson stat1.581933 Prob(F-statistic)0.002654Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:52Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C126.302979.749531.5837440.1356X20.2501510.0358196.9838280.0000R-squared0.776977 Mean dependent var585.6875Adjusted R-squared0.761047 S.D. dependent var368.9720S.E. of regression180.3639 Akaike info criterion13.34430Sum squared resid455436.0 Schwarz criterion13.44087Log likelihood-104.7544 F-statistic48.77385Durbin-Watson stat2.702078 Prob(F-statistic)0.000006Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:49Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C387.9315140.39822.7630800.0152X30.9434050.5287381.7842580.0961R-squared0.185269 Mean dependent var585.6875Adjusted R-squared0.127073 S.D. dependent var368.9720S.E. of regression344.7325 Akaike info criterion14.63988Sum squared resid1663767. Schwarz criterion14.73646Log likelihood-115.1191 F-statistic3.183575Durbin-Watson stat2.112477 Prob(F-statistic)0.096062Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:50Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C104.7661310.11510.3378300.7405X40.1603170.0991631.6166990.1282R-squared0.157323 Mean dependent var585.6875Adjusted R-squared0.097132 S.D. dependent var368.9720S.E. of regression350.5949 Akaike info criterion14.67361Sum squared resid1720835. Schwarz criterion14.77018Log likelihood-115.3889 F-statistic2.613717Durbin-Watson stat1.642909 Prob(F-statistic)0.128246Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:50Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C239.787360.274603.9782480.0014X50.1720910.0219357.8456600.0000R-squared0.814703 Mean dependent var585.6875Adjusted R-squared0.801467 S.D. dependent var368.9720S.E. of regression164.4028 Akaike info criterion13.15898Sum squared resid378396.0 Schwarz criterion13.25556Log likelihood-103.2719 F-statistic61.55438Durbin-Watson stat2.497236 Prob(F-statistic)0.000002综合比较检验和T检验发现的拟合效果较好，从而得到基本方程：逐一引入其他变量：Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:58Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C151.790389.060001.7043600.1121X50.1481240.0280985.2716270.0002X10.1056250.0803271.3149300.2113R-squared0.836455 Mean dependent var585.6875Adjusted R-squared0.811294 S.D. dependent var368.9720S.E. of regression160.2825 Akaike info criterion13.15911Sum squared resid333976.1 Schwarz criterion13.30397Log likelihood-102.2729 F-statistic33.24441Durbin-Watson stat2.614918 Prob(F-statistic)0.000008Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:58Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C170.628071.203472.3963430.0323X50.1078290.0447122.4116460.0314X20.1079950.0665521.6227070.1286R-squared0.845914 Mean dependent var585.6875Adjusted R-squared0.822208 S.D. dependent var368.9720S.E. of regression155.5785 Akaike info criterion13.09954Sum squared resid314660.8 Schwarz criterion13.24440Log likelihood-101.7963 F-statistic35.68411Durbin-Watson stat2.549746 Prob(F-statistic)0.000005Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 19:59Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C224.615473.256423.0661530.0090X50.1678640.0250706.6958590.0000X30.1129100.2881980.3917810.7016R-squared0.816865 Mean dependent var585.6875Adjusted R-squared0.788691 S.D. dependent var368.9720S.E. of regression169.6105 Akaike info criterion13.27225Sum squared resid373980.3 Schwarz criterion13.41711Log likelihood-103.1780 F-statistic28.99300Durbin-Watson stat2.538863 Prob(F-statistic)0.000016Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:00Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C382.9909150.12742.5511060.0241X50.1870790.0261847.1447680.0000X4-0.0577800.055509-1.0409160.3169R-squared0.828959 Mean dependent var585.6875Adjusted R-squared0.802645 S.D. dependent var368.9720S.E. of regression163.9147 Akaike info criterion13.20393Sum squared resid349284.3 Schwarz criterion13.34879Log likelihood-102.6314 F-statistic31.50252Durbin-Watson stat2.894780 Prob(F-statistic)0.000010保留继续引入变量Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:16Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C130.305289.822581.4506960.1725X50.1062440.0455092.3345630.0378X20.0856560.0738001.1606410.2684X10.0655720.0864610.7583960.4628R-squared0.852961 Mean dependent var585.6875Adjusted R-squared0.816201 S.D. dependent var368.9720S.E. of regression158.1847 Akaike info criterion13.17772Sum squared resid300268.8 Schwarz criterion13.37087Log likelihood-101.4218 F-statistic23.20370Durbin-Watson stat2.717507 Prob(F-statistic)0.000028Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:16Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C162.225480.792572.0079240.0677X50.1062280.0468252.2686400.0425X20.1062190.0694231.5300350.1519X30.0709950.2757570.2574550.8012R-squared0.846760 Mean dependent var585.6875Adjusted R-squared0.808450 S.D. dependent var368.9720S.E. of regression161.4859 Akaike info criterion13.21903Sum squared resid312932.3 Schwarz criterion13.41218Log likelihood-101.7522 F-statistic22.10284Durbin-Watson stat2.578380 Prob(F-statistic)0.000035Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:17Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C289.1764157.90951.8312800.0920X50.1249910.0495832.5208680.0269X20.0989920.0681441.4526790.1720X4-0.0455060.053946-0.8435440.4154R-squared0.854539 Mean dependent var585.6875Adjusted R-squared0.818174 S.D. dependent var368.9720S.E. of regression157.3337 Akaike info criterion13.16693Sum squared resid297046.7 Schwarz criterion13.36008Log likelihood-101.3355 F-statistic23.49877Durbin-Watson stat2.898481 Prob(F-statistic)0.000026保留继续引入变量Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:20Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C243.2397175.89071.3829030.1941X50.1221530.0509592.3971090.0354X20.0797030.0755911.0544010.3143X4-0.0417520.055536-0.7517930.4680X10.0587980.0885310.6641480.5203R-squared0.860147 Mean dependent var585.6875Adjusted R-squared0.809291 S.D. dependent var368.9720S.E. of regression161.1308 Akaike info criterion13.25262Sum squared resid285594.6 Schwarz criterion13.49405Log likelihood-101.0209 F-statistic16.91350Durbin-Watson stat3.062277 Prob(F-statistic)0.000115Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:21Sample: 1 16Included observations: 16VariableCoefficientStd. Errort-StatisticProb. C280.5572169.97501.6505790.1271X50.1234330.0522532.3622400.0377X20.0977030.0713221.3698960.1980X4-0.0447940.056348-0.7949520.4435X30.0571480.2806260.2036430.8424R-squared0.855085 Mean dependent var585.6875Adjusted R-squared0.802389 S.D. dependent var368.9720S.E. of regression164.0208 Akaike info criterion13.28817Sum squared resid295931.1 Schwarz criterion13.52960Log likelihood-101.3054 F-statistic16.22668Durbin-Watson stat2.928913 Prob(F-statistic)0.000139综合比较后发现，引入后，使拟合优度提高，但是对的参数值有明显的影响，且统计检验也不显著，由此可以断定之间存在共线性，舍弃，其他变量对模型的影响均不显著，均舍弃。最后得到修正后的模型：2对随机误差项之间异方差性的检验：（1）图示法：由图可见其异方差性是相当明显的。（2）GoldfeldQuandt检验：Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 20:59Sample: 1 6Included observations: 6VariableCoefficientStd. Errort-StatisticProb. C78.9739150.633621.5597130.1938X50.2959290.0660864.4779440.0110R-squared0.833694 Mean dependent var274.5000Adjusted R-squared0.792117 S.D. dependent var137.7252S.E. of regression62.79473 Akaike info criterion11.37882Sum squared resid15772.71 Schwarz criterion11.30941Log likelihood-32.13646 F-statistic20.05198Durbin-Watson stat2.910560 Prob(F-statistic)0.011007Dependent Variable: YMethod: Least SquaresDate: 05/17/05 Time: 21:01Sample: 11 16Included observations: 6VariableCoefficientStd. Errort-StatisticProb. C344.0503150.85182.2807180.0847X50.1460040.0358774.0695510.0152R-squared0.805459 Mean dependent var867.3667Adjusted R-squared0.756824 S.D. dependent var391.7530S.E. of regression193.1847 Akaike info criterion13.62637Sum squared resid149281.3 Schwarz criterion13.55696Log likelihood-38.87912 F-statistic16.56124Durbin-Watson stat2.868491 Prob(F-statistic)