《多元回归与多重共线数据.doc》由会员分享，可在线阅读，更多相关《多元回归与多重共线数据.doc（6页珍藏版）》请在金锄头文库上搜索。
1、实验一：多元回归【实验目的】掌握建立多元回归模型和比较、筛选模型的方法。【实验内容】建立我国国有独立核算工业企业生产函数。根据生产函数理论，生产函数的基本形式为：。其中，L、K分别为生产过程中投入的劳动与资金，时间变量反映技术进步的影响。表1列出了我国1978-1994年期间国有独立核算工业企业的有关统计资料；其中产出Y为工业总产值（可比价），L、K分别为年末职工人数和固定资产净值（可比价）。请比较、分析建立合适的回归模型。表1 我国国有独立核算工业企业统计资料年份时间工业总产值Y（亿元）职工人数L（万人）固定资产K（亿元）197813289.1831392225.70197923581.26
2、32082376.34198033782.1733342522.81198143877.8634882700.90198254151.2535822902.19198364541.0536323141.76198474946.1136693350.95198585586.1438153835.79198695931.3639554302.251987106601.6040864786.051988117434.0642295251.901989127721.0142735808.711990137949.5543646365.791991148634.8044727071.3519921597
3、05.5245217757.2519931610261.6544988628.7719941710928.6645459374.34资料来源：根据中国统计年鉴1995和中国工业经济年鉴-1995计算整理第一步：Y=a+bL+cK第二步：Y=aLbKcLnY=a+blnL+clnKDependent Variable: YMethod: Least SquaresDate: 11/05/10 Time: 10:24Sample: 1978 1994Included observations: 17VariableCoefficientStd. Errort-StatisticProb.C-238
4、7.269816.8895-2.9223900.0111L1.2085320.2730204.4265280.0006K0.8344960.05742114.532870.0000R-squared0.995617Mean dependent var6407.249Adjusted R-squared0.994990S.D. dependent var2486.742S.E. of regression176.0069Akaike info criterion13.33771Sum squared resid433697.8Schwarz criterion13.48475Log likeli
5、hood-110.3705F-statistic1589.953Durbin-Watson stat1.481994Prob(F-statistic)0.000000Dependent Variable: LNYMethod: Least SquaresDate: 11/05/10 Time: 10:28Sample: 1978 1994Included observations: 17VariableCoefficientStd. Errort-StatisticProb.C-1.9512531.665320-1.1716980.2609LNL0.6044670.2726972.216625
6、0.0437LNK0.6736580.0723579.3101300.0000R-squared0.995753Mean dependent var8.692837Adjusted R-squared0.995147S.D. dependent var0.394921S.E. of regression0.027512Akaike info criterion-4.189602Sum squared resid0.010597Schwarz criterion-4.042564Log likelihood38.61162F-statistic1641.407Durbin-Watson stat
7、1.338201Prob(F-statistic)0.000000实验二：多重共线性【实验目的】掌握多重共线性的检验及处理方法【实验内容】建立并检验我国钢材产量预测模型【例2】表2是19781997年我国钢材产量（万吨）、生铁产量（万吨）、发电量（亿千瓦时）、固定资产投资（亿元）、国内生产总值（亿元）、铁路运输量（万吨）的统计资料。一、检验多重共线性二、利用逐步回归方法处理多重共线性表2 我国钢材产量及其它相关经济变量统计资料年份钢材产量Y生铁产量X1发电量X2固定资产投资X3国内生产总值X4铁路运输量X51978220834792566668.723264110119197924973673
8、2820699.3640381118931980271638023006746.945181112791981267034173093638.2148621076731982292035513277805.952951134951983307237383514885.26593511878419843372400137701052.43717112407419853693438441071523.51896413070919864058506444951795.321020213563519874386550349732101.691196314065319884689570454522554
9、.861492814494819894859582058482340.5216909151489199051536238621225341854815068119915638676567753139.032161815289319926697758975394473.762663815762719937716895683956811.353463416266319948428974192819355.354675916309319958980105291007010702.975847816585519969338107231081312185.796788516880319979979115
10、111135613838.9674463169734一、检验多重共线性相关系数检验利用相关系数可以分析解释变量之间的两两相关情况。在Eviews软件中可以直接计算相关系数矩阵。本例中，在Eviews软件命令窗口中键入：COR X1 X2 X3 X4 X5辅助回归方程检验当解释变量多余两个且变量之间呈现出较复杂的相关关系时，可以通过建立辅助回归模型来检验多重共线性。本例中，在Eviews软件命令窗口中键入：LS X1 C X2 X3 X4 X5LS X2 C X1 X3 X4 X5LS X3 C X1 X2 X4 X5LS X4 C X1 X2 X3 X5LS X5 C X1 X2 X3 X4二
11、、利用逐步回归方法处理多重共线性建立基本的一元回归方程逐步引入其它变量，确定最适合的多元回归方程C-567.9033109.5040-5.186143X10.9214020.01622056.80710C-34.5711791.77101-0.376711X20.8841060.01414962.48594C2896.257211.002213.72619X30.5725000.03697915.48188C2720.437203.490413.36887X40.1088420.00651716.70220C-9760.8511317.466-7.408806X50.1068320.00932
12、711.45381C309.6478933.55410.3316870.7447X20.7837950.2125553.6874870.0022X30.5337170.1248514.2748430.0007X4-0.0851990.027765-3.0685650.0078X50.0078840.0128150.6152050.5476C-287.6867101.2341-2.841797X20.4871850.1126874.323352X10.4158670.1174973.539376C200.8050168.49481.191758X20.8085640.04810316.80884X30.0527340.0322241.636499C
