《练习题及参考解答第四版计量经济学.docx》由会员分享,可在线阅读,更多相关《练习题及参考解答第四版计量经济学.docx(14页珍藏版)》请在三一办公上搜索。
1、第四章练习题及参考解答4假设在模型y铲+OX+Pj师中1XBX之间的相关系数为零,有人建议你分别进行如下回归:y=y+.x+uI33/Z1.(1)是否存在戊=P且V=B?为什么?2233人(2)%会等于4或或者两者的某个线性组合吗?(3)是否有V3)=Var(y)?【练习题4.1参考解答】存在戊=8且V=62233因为同理有:(Xyx)(KJ-)(1)2)2)-Cx)2123/之间的相关系数为蹲时,离差形式的ZV3。(Xyx)(4x2)S%22/3i2Y=833Q)会的。Q)=var(d)且varQ)=var存在Var02因为var=Zx2G-n)3F=UUI-(p)23叼,varx7,2/2
2、3varQ)=var(y)同理,有33022/4,2表4.4给出了19952016年中国商品进口额Y、国内生产总值GDP、居民消费价格指数CP1.的数据。表4.4中国商品进口额、国内生产总值、居民消费价格指数年份商品进口额(亿元)Y国内生产总值(亿元)GDP居民消费价格指数(1978=100)CPI199511048.161339.9396.9199611557.471813.6429.9199711806.579715.0441.9199811626.185195.5438.4199913736.490564.4432.2200018638.8100280.1434.0200120159.2
3、110863.1437.0200224430.3121717.4433.5200334195.6137422.0438.7200446435.8161840.2455.8200554273.7187318.9464.0200663376.9219438.5471.0200773284.6270232.3493.6200879526.5319515.5522.7200968618.4349081.4519.0201094699.3413030.3536.12011113161.4489300.6565.02012114801.0540367.4579.72013121037.5595244.45
4、94.82014120358.0643974.0606.72015104336.1689052.1615.22016104967.2744127.2627.5资料来源:中国统计年鉴2017考虑建立模型:Iny=P+PInGDP+PInCP1.+u(1)利用表中数据估计此模型的参数。(2)你认为数据中有多重共线性吗?(3)进行以下回归:InY=A+AInGDP+v112tInY=+21.nCP/+V2;InGDP=C1.+C2InCP/+13根据这些回归你能对多重共线性的性质有什么认识?(4)假设经检验数据有多重共线性,但模型中B和8在-5%水平上显著,并且F检验也显23著,你对此模型的应用有何
5、建议?【练习题4.2参考解答】建立模型:InY=9+O2InGDP+PInCP1.+u(1)利用表中数据估计此模型的参数。=JUHTT1.EDWOftIin341.M/WIUrKVmRnToo1.H1PnrdNEC.FgzIEuxwrntoreeiIISdt投S由IDEaHId1.rvIVInJbIr1.HYM1.CM1.Hdia0HEIIff1.1W52G13IndudGdOHimJimi1.Ufmd3bM转!ShJQffW1SMWPTOPt,J.!”M4E1外i1.IJ_iQVUK)OP1.NCRZOOBUaQDJB8Q412.M45Tftjoxka收徽里白之岳8mRW 0U4TM1SX
6、.rrg,3jSue明/曲门I】mid1.即解防翼煤 F PnmPtfS*osooNB681&411B4ZU1.Q49092*USM73叫49Utdn23A,KIIcn92I57-SO,AMrzItwncE,twsHtnMIIHQunn51削O-XvOSwni1.1.1CMMIXKO(2)你认为数据中有多重共线性吗?其中居民消费价格指数CP1.对商品进口额影响为负,与预期不符合,可能存在多重共线性。(3)分别进行以下回归:1)作回归Iny=A+AInGDP+v/12/If11EIa:UWTIT1.F11wtte.RJUrff而出V(UM.RintrwnCFrteteEi1.eMIFOeusCS
7、trtpRendaVoiiVIt.1.MYU1.cfted1.ttttSq1.iart)&a1.QWiAi8T1.mIg:43SampHi:austtJ.19f5HJ13Vanab1.eJOmQ电时SUEtfOrk31.afeE4cProbC31678Q7B4gT65S472QKQ71.NSF19,忡.t.12465口刖如R6制.田O时1仅57,3和,I1.H S.E. cA regression Sum squared red Icig* kt1t4 4 FiIakoAcPr&b仔 5MsCc的(FU*andPAfXWYW3rQ947W? 30 depeMerrM 02D272Sift Cf
8、TIertCnMBMT5SdiwaizcfiUenion187A8 Hannan-au1.n carter WitDurbtfvWafcsun staffWSMID&801M-0JMS05 41S51MIII 237TMQ第如R?说明GDP的确对商品进口额有正的影响,是重要变量。2)作回归Iny=BB21.nCPI+vnMEqum度UNTrnEPKOrR加:醉/M4图至喊足口。Oei:-endArYIV31.aE1.:1.MMrtwa1.HT那装.i03JW18T1.nw18Samo1.f1.1995201Sfeidu1.MdgwsvgibIfi322mCMaanS9f1.M0417734心北
9、tomfoenttnan11打5联BumSqUIrtdBeS汨M4的QTSchwarZCAmcki12H74B1.ogMWIWSrfmi9HamSMQUinn5)if)1201527FdHf阳匕78735J9DurtHH-WstecnSis1.O1W515Frob(F-s1.arti51.!cOKKKKK)说明CP1.的确对商品进口额有正的影响,是重要变量。3)作回归InGDP=C1.+C2InCP1.+v3,DtPUndtniVanabit1.MCDPMeWiQd1.etSquares.报口出,口用Timt13J4variabiSMEHMkM昭PHI&C2542457K5-10.860170
10、00001.NcF1.o*1.ff1.3?O.I7*7S516.03007D.OQQOUi1.i1.o*11OtnincKted11II0F*19afterav*etsR-war11dMvuff1.TOKdQuarodQrreQfeSSIOdSumsuaredres-k1.1.gIIKthd;F1.ja抑CFro&fF-siI1.1.1.uUearidependenrtvacSOdspedeCEOn0MW2FSdiwarzqoo5.723313Hannan-KJuinnc24428538721D95-2S010S5O.D15D700943410.0a04763&4139ODOOD&SZE1.U
11、24&790.03656910,60990OOOODGYZJZ-O.IESB,BHQ.O215S07.745a330.D0RsquaredAdjustedRsquaredS.E.afrearessionSumsquaredresid1.ogSkeihoodF-StaI1.StIcFrobtF-Statistic)O.999aS609998606R1.64554631679.-130.360638101.2200D00Meandependenivar3.DdependentAarAkaikeinfocterianSchwarzcntGriDnHanan-aumnenter.OUrbin-Wats
12、cInstat725S1735153142I5.69616.0620115.3EM2618160计算解释变量的相关系数:Corre1.ationGDPGYZJZSSZEGDP1.000(X)00.992504-0.998301GYZJZ0.9925041.0000000.995185SSZE0.998S010.9951851.003000解释变量的方差扩大因子VIF:VarianceInf1.ationFactorsDate:()3)13J18Time:04:15Samp1.e:20002016Inc1.ud1.edobservations:T7Variab1.eCoef1.cientVari
13、anceUncenteredVIFGenIeredVIFC7605-74.93473354NAGDP0.0002201756.769461.6316SSZE0.00932623-79.835717.6818GYZJZ0000435509.2540115.1571这说明由于严重的多重共线性导致工业增加值的参数为负。工业增加值与国内生产总值、税收总额都高度相关,为分析工业增加值对一般公共预算收入是否有负的影响,可删除国内生产总值、税收总额作回归:u4M NY M塔 big CZSR Method: 1.eaail Souares Date 03/17/16 Time: 11:29SWENW 200
14、02010 Indudid QSservjijons 17Var1.abMCoeffiasfliSid.Errorl-SiabsbcPfO&.GYZJZ-19103 S4 0.4735453524.4D35J21BB70.Q2277329.57607O.DOT10.0000R-squared .MjuAtedRAquar ed; SE VfirtgireaisAan Sum squared resid Log Ilkalihwd F,SWISIICO.S82O1B 6910159 7.16EM 173.3508 174 7913 O.MMODKfeara dependent var S.D d
15、-aDinn otter.DurDin-WaIson Stit72591.73 5153142 20 02950 20,72753 20,53925 0.2787B1这说工业增加值对一般公共预算收入是有证的租金作用的。不过国内生产总值、税收总额都是时一般公共预算收入有重要影响的变量,删除后可能模型会有设定误差。4.4 表4.6是中国家电零售总额及国内生产总值、人均可支配收入、家电广告投放总额、居民消费价格指数等数据。表4.6997年一2015年中国家电零售F总额及相关数据年份家电零售总额n(亿元)GDPX2(亿元)人均可支配收入X3(元)家电广告投放总额X4(亿元)居民消费价格指数XS(以19
16、96年为100)1997506.078802.95160.364.71102.81998651.783817.65425.179.02102.01999724.389366.55854.067.14100.52000831.699066.16280.073.51101.02001784.7109276.26859.665.88101.72002953.0120480.47702.878.74100.820031127.2136576.38472.288.00102.120041415.7161415.49421.676.51106.020051636.0185998.910493.077.41
17、07.920061921.7219028.511759.588.61109.620072370.727084413785.894.40114.820082706.6321500.515780.887.92121.620093154.4348498.517174.798.67120.720104056.5411265.219109.4119.43124.720115374.9484753.221809.8140.34131.520125935.8539116.524564.7205.09134.920136944.5590422.426955.1229.73138.420147603.36447
18、91.129381.0246.83141.220158269.5682635.131790.3277.19143.1数据来源:国家统计局(WWW.1.(1)如果请考虑建立模型:t=O+022,+P3X3,+P44+PsXS+%,利用表中数据估计此模型的参数。(2)根据模型估计结果,你认为参数估计结果合理吗?数据中有吗?(3)分别采用简单相关系数检验法和方差扩大因子法验证模型是否存在多.重共线性。(4)如果存在多重共线性,如何才能解决?【练习题4.4参考解答】O1.S方法估计模型参数,得到的回归结果。DependentVariab1.e:YMethod:1.eastSquaresDais:037
19、13;18Time:02:11Samp1.e:19972015Inc1.udedabsenjations:19Variab1.eCoef1.icientStd.Error(-StatisticProb.C30777852926.1991D518050.2107X20.0207420.0058823.5263450.0024.X3-0.2117210.10294Q2.056654O,O6BgX410,3919215921806.0400340.0000X5-37.63322291883-12792200.2216R-squared0.997674Meandependentvar2993321Ad
20、justedR-squared0.997009S.D.dependentvar2533.623S.E.ofregression141.2925Akaikeinfccriterion1296048Suitsquaredresid2794S0.0Schwarzcriterion13,209011.og1.ike1.ihood-110.1245Hannan-Quinnenter.13,00254F-statistic1501140Jr1.in-Watsonstat2.665752Prot(F-Statistic)0.000000该模型R2=0.9977,R2=0.9970,可决系数很高,F检验值15
21、01.140,明显显著。但是当a=0.05时1.U=25a9-5)=2.145,不仅X3、X5的系数不显著,而0X3的符号与预期相反,这表明可能存在严重的多重共线性。计算各解释变量的相关系数:变量X2X3X4X5X21.0000000.9989320.9375850.995145X30.9989321.0000000.9412570.991385X40.9375850.9412571.0000000.917290X50.9951450.9913850.9172901.000000由相关系数矩阵可以看出,所有解释变量之间的相关系数较高,证实确实存在一定的多重共线性。解释变量的方差扩大因子V1.F
22、被解释变量方差扩大因子X21335.613X3723.4567X410.4945X5185.1135所有解释变量的方差扩大因子都远大于10,表明存在严重多重共线性问题。对多重共线性的处理:将各变曷进行对数变换(x5除外),再对以下模型进行估计.Iny=o+Q1.nX2+0nX3+小X4+PsX5+wDependentVariab1.e:1.NYMethod:1.eastSquaresDate:0.3/13/18Time:0217samp1.e:19972015Inc1.udedobservations:19Variab1.eCoefficientStdError(-StatisticProbC
23、-7.6950870.6977D8-11.029100.00001.NX21.335317O.5478D12.4375940.02871.NX30.1348170.673249-0.2745140.78771.NX40.32D4020.1205792491g560.0259)(5-0.0077420.005346-1.4481210.1696R-SquarDd0.997446Meandependentvar7.626478AdjustedR-squared0.996716S.D.dependetvar0.91B548S.E.ofregression0.052639Akaikeii1.foCri
24、terior2S297S7Sumsquaredresid0.03B792Sctiwarzcriterion-25812511.og1.ike1.ihood31.8B29SHannan-Ouinncfite.2787725F-statistic1366.756Durbin-Watsonsta11.624259ProbiF-Statistic)0.00D000该模型R2=0.9974,R2=0.9967,M决系数很高,F检验值1366.756,明显显著。但是取a=0.05时)(-A)=1-2.145,1.NX3和1.NX5依然不显著,且1.NX3的回归系数符号与预期不相符,这表明经对数变换后的模型
25、依然存在严重的多重共线性,需要更进一步的修正方法。采用逐步回归方法筛选并剔除引起多重共线性的变量,最后保留的解释变量为乂和X,估计结果为人Y=-958.3489+0.0090X2+10.9547X4(75.0751)(0.0005)(1.5102)t=(-12.7652)(18.2819)(7.2538)Ri=0.9970Ri=0.9966f=2627.512该模型中H2=0.9970,R2=0.9966,可决系数很高,F检验值2627.512,明显显著。当a=0.05时tn-k)=t(19-5)=2.145,所有系数估计值高度显著。对系数估计值a120.025的解释如下:在其他变量保持不变的
26、情况下,如果国内生产总值GDP每增加一亿元,则家电零售总额将增加90万元;家电广告投放总额每增加一亿元,则家电零售总额将增加10.9547亿元。4.5 表4.7中给出了四川省城镇人均消费支出Y和其相关影响因素城镇居民人均可支配收入死、地区生产总值X3、零售商品价格指数X4以及人口自然增长率X5的数据。4.72000-2015城镇居民人均消费支出及其影响因素数据城镇人均消费支出(元)Y城镇居民人均可支配收入(元)X2地区生产总值(亿元)X3零售商品价格指数(%)X4人口自然增长率(%。)X5200053156360.53928.297.75.1200156615894.34293.49100.8
27、4.37200259326610.84725.0199.43.89200363127041.95333.09100.13.12200469707709.96379.63103.72.782005757783867385.1100.62.9200683059350.18690.24101.72.862007955911098.310562.39105.32.9220081060812633.412601.23105.32.3920091170113839.414151.28100.12.7220101345715461.217185.481032.3120111568717899.121026.
28、68104.62.982012166492030723872.8101.62.972013178992236826392.07101.732014193182423428536.66100.63.22015201142620530053.1100.23.36(数据来源:四川统计局,国家统计局)(1)利用以上数据建立回归模型,根据回归结果判断是否存在严重的多重共线性。(2)若对所建立的模型中部分变量作对数变换(某些变量不需要进行对数变换),并对变换后的模型进行估计,根据同归结果判断是否还存在严重的多.重共线性。(3)若是还存在严重的多重共线性,选择更适合的模型进行修正,并对修正后的结果从经济意义
29、上进行解读。【练习题4.5参考解答】(1)回归模型估计结果为:DependentVariab1.e:YMethod:1.east&quaresDate:03/13J18Time:03:45Samp-Ie:20002015Inc1.udedobservations:16Variab1.eCoefficientStd.Errort-StaiistieProb.CB18.8E8O3S97.9310.210073G.S374-X2-0.0500780.150123-0.333567Ci.7450X30.5-937230.1113075.3102400.0002X4323917634,639360.93
30、5IU0.3698X&-149.20211071556-1.3923870.191JK-squared0.998S11Meandependentvar11316.50AdjuatedR-squaird0.998379S.D.dependentvar5193.535-S.E.ofregression209.1036Akaikeinfocriterion13,77334Sumsqua期resid480967.4Schwarzcriterion14,015281.og1.ike1.ihood-105.1907Hannan-C1.uinncriter.13,78621F-statistic231056
31、3urbin-Watsonstat1.470036Prob(F-statisiic;0.000000结果中的m=0.9988,R=0.9984,非常高,F统计量为2310.563,非常显著。但是X2,X4,X5的t统计量并不显著,且X2的系数为负,与实际不符。因此可能存在严重的多重共线性。计算解释变量的相关系数:Corre1.ationX2X3X4X5X21.0000000.9984490.1200+S-Q.326115X303964491.0000000.138862-0.338351X40.1200460.13BS621.000000-0.669248X5-032S115-0.33B361
32、-0.6692481.000000计算方差扩大因子:VarianceInf1.ation Factors ate: 03dS18Time: 03:47Sample: 2000 2015Included observations: 16Variab1.eCoefficientVarianceUncenteredVIFCenteredVIFC151938675559.879NAX20.0225331858.826364.0 793X30.0125011271.532365.9054X41199 s35453s 7921.372535X511402.3344,510742.037964(2)由于X4
33、和X5一经是指数和比率,因此只对Y和X2,X3做对数变换,估计结果DependentVariab1.e:1.OG(Y)Method:1.east&quaresDa1.e:0313J18Time:03:46Samp-Ie:20002015Inc1.udedObservaiions:16Variab1.eCoefficien1.StdErrort-StaiisticProbC2.9265240.5579115.2455010.00031.OGCX2)O.OO7B3701599030.0493240.96151.OGX3)0.6657250.1199665.5492970.0002X4-0.0007
34、990.003383-0.2767060.7871X50.0329430.0124762.6404220.0230R.-squared0.999049Meandependentvar9.232829AdjustedR-squared0.998703S.D.dependentvar0.467S415.E.ofregression0.016S5-1.Akaikeinfocriterion-5.07S465Sumsquaredresid0.003124Schwarzcriterion-4.8370311.eg1.ike1.ihood45,62772Hannan-Ciuinncriter.-5.066
35、101F-statistic837,656Durbin-Watsonstat1.153032Prob(F-Statistic)0.000000虽然Iog(X2)的系数为正,但是依然不显著,说明严重的多重共线性依然存在。计算变换后的解释变量的相关系数:1.OGQC2LC1G(X3X4X5LOG(X2)1.0000000.9951760.241612-0.460S30LOG,(X3)0.9951761.0000000.306446-0.531069X4024161203064461.000000-0.66924SX5-0X60380-0.531069-0.6692481.000000Corre1.
36、ation计算方差扩大因子:VariantInf1.ation Factors Date: 03J1318 Tione: 03:46Sample: 2000 2015 iddedo bE9rVHi ons: 16Variab1.eCaeHcIentVarianceUncentored1.VIFCamaredMFC0.31126517538.02NA1.OGCK2)O.D2S569127245.6356.1250ICEXaO.D143927D921.383S9.0SX49.a4EO64B59.2D62.111795K5O.DD015092911694.254045(3)对(1)中回归模型的多重共线性进行修正,由于(2)的对数变换,依然存在严重的多重共线性,所以进一步对(1)的回归模型采用逐步回归方法进行修正:分别作丫对X2、X3、X4、Xf的回归,结果为:变量X2X3X4X5参数估计值0.75430.5616394.2835-2639.516t统计量45.657680.04950.6306-1.4695R