計(jì)量經(jīng)濟(jì)學(xué)課程設(shè)計(jì)論文

1、計(jì)量經(jīng)濟(jì)學(xué)課程設(shè)計(jì)第 頁(yè)共28頁(yè)目錄 TOC o 1-5 h z HYPERLINK l bookmark2 o Current Document 1引言 2 HYPERLINK l bookmark4 o Current Document 1 .1 金融業(yè)簡(jiǎn)介2 HYPERLINK l bookmark6 o Current Document 1 .2課題的意義2 HYPERLINK l bookmark8 o Current Document .3課題的內(nèi)容和任務(wù)2 HYPERLINK l bookmark10 o Current Document 2建立模型及多元回歸 2. 1模型建立

2、3 HYPERLINK l bookmark12 o Current Document 2 . 2多元回歸 4 HYPERLINK l bookmark24 o Current Document 3回歸模型的檢驗(yàn)和預(yù)測(cè) 5 HYPERLINK l bookmark26 o Current Document 經(jīng)濟(jì)意義上變量的檢驗(yàn) 5 HYPERLINK l bookmark28 o Current Document 3.1.1擬合優(yōu)度檢驗(yàn) 5 HYPERLINK l bookmark32 o Current Document 3.1.2對(duì)方程的T檢驗(yàn) 53.1.3 對(duì)方程的 F檢驗(yàn) 6 HYPE

3、RLINK l bookmark50 o Current Document 3.2計(jì)量經(jīng)濟(jì)學(xué)檢驗(yàn) 63.2.1多重共線性檢驗(yàn) 6 HYPERLINK l bookmark70 o Current Document 3.2.2異方差性檢驗(yàn) 15 HYPERLINK l bookmark120 o Current Document 3.2.3序列相關(guān)性檢驗(yàn) 20 HYPERLINK l bookmark158 o Current Document 3.3滯后變量模型 26 HYPERLINK l bookmark174 o Current Document 3.4模型參數(shù)的檢驗(yàn) 27 HYPERL

4、INK l bookmark176 o Current Document 點(diǎn)估計(jì) 27 HYPERLINK l bookmark182 o Current Document 3.4.2區(qū)間估計(jì) 27 HYPERLINK l bookmark196 o Current Document 3.5模型的預(yù)測(cè) 2829 HYPERLINK l bookmark198 o Current Document 結(jié)論 29參考文獻(xiàn)1引言1.1金融業(yè)簡(jiǎn)介金融業(yè)指的是銀行與相關(guān)資金合作社， 還有保險(xiǎn)業(yè)，除了工業(yè)性的經(jīng)濟(jì)行為 外，其他的與經(jīng)濟(jì)相關(guān)的都是金融業(yè)。 金融業(yè)是指經(jīng)營(yíng)金融商品的特殊行業(yè)， 它 包括銀行業(yè)、保

5、險(xiǎn)業(yè)、信托業(yè)、證券業(yè)和租賃業(yè)。隨著WTO來(lái)臨以及IT技術(shù)的日益精進(jìn)，金融業(yè)已是二十一世紀(jì)的朝陽(yáng)行業(yè)，在國(guó)家整個(gè)國(guó)民經(jīng)濟(jì)中處于牽一 發(fā)而動(dòng)全身的地位，行業(yè)運(yùn)轉(zhuǎn)是否良好， 事關(guān)經(jīng)濟(jì)發(fā)展和社會(huì)穩(wěn)定，具有優(yōu)化資 金配置和調(diào)節(jié)、反映、監(jiān)督經(jīng)濟(jì)的作用。1.2課題的意義從整個(gè)經(jīng)濟(jì)市場(chǎng)的發(fā)展趨勢(shì)出發(fā)， 考察了貨幣供應(yīng)量、上市公司數(shù)量、股票 的發(fā)行量、股票籌資額等對(duì)我國(guó)金融業(yè)的影響。因此，通過(guò)研究這些貨幣供應(yīng)量、上市公司數(shù)量、股票的發(fā)行量、股票籌資 額等對(duì)我國(guó)金融業(yè)的影響，推動(dòng)我國(guó)金融業(yè)的一體化、綜合化、網(wǎng)絡(luò)化發(fā)展的進(jìn)程，優(yōu)化了金融資源的合理配置，提高了競(jìng)爭(zhēng)能力，改進(jìn)了金融服務(wù)，降低了經(jīng) 營(yíng)成本，增加了盈利，對(duì)

6、發(fā)展中國(guó)金融業(yè)的競(jìng)爭(zhēng)實(shí)力有很大的幫助。1.3課題內(nèi)容和任務(wù)本文將在經(jīng)濟(jì)理論的指導(dǎo)下，采用計(jì)量經(jīng)濟(jì)的方法，并借助于計(jì)量經(jīng)濟(jì)學(xué)軟 件Eviews，對(duì)我國(guó)的貨幣供應(yīng)量、上市公司數(shù)量、股票的發(fā)行量和股票籌資額 等對(duì)金融業(yè)進(jìn)行初步的實(shí)證分析。2建立模型及多元回歸表一 1993年到2010年金融業(yè)及其影響因素的數(shù)據(jù)統(tǒng)計(jì) 上市金融業(yè)總 資產(chǎn)（億貨幣供應(yīng)量（億元）黃金儲(chǔ) 備（萬(wàn) 盎司）外匯儲(chǔ)備（億美公司數(shù)量（個(gè)股票的發(fā) 行量（億股票籌資額（億元）年份元）丫X1X2元）X3)X4股）X5X619931669.74534879.81267211.9918395.79375.4719942234.84446923.

7、51267516.229191.26326.7819952798.50360750.51267735.9732331.6150.3219963211.68576094.912671050.2953086.11425.0819973606.76290995.312671398.9745267.631293.8219983697.667104498.512671449.59851109.06841.5219993816.459119897.912671546.75949122.93944.5620004086.686134610.312671655.741088512.042103.2420014

8、353.456158301.916082121.651160141.481252.3420024612.80118500719292864.071224291.74961.7520034989.396221222.819294032.511287281.431357.7520045392.97525410719296099.321377227.921510.9420056086.826298755.719298188.721381567.051882.5120068099.082345603.6192910663.414341287.775594.29200712337.55403442.21

9、92915282.491550637.2409868025475166.6192919460.31625180.293852.21200917767.53606225338923991.521718415.966124.69201020980.63725774.1338928473.382063928.3711971.932.1模型建立根據(jù)統(tǒng)計(jì)數(shù)據(jù)表一金融業(yè)總資產(chǎn)模型，首先建立下面的模型：YXX X、 X 5 X XYt01X1t 2X2t 3X3t 4X4t 5X5t 6X6tt其中Yt是金融業(yè)總資產(chǎn)，Xit是貨幣供應(yīng)量，X2t是黃金儲(chǔ)備，X3t是外匯儲(chǔ)備，X4t是

10、上市公司數(shù)量，X5t是股票的發(fā)行量，X6t是股票籌資額，S是常數(shù)項(xiàng)， (j =1,2,3)是待估參數(shù)，隨機(jī)干擾項(xiàng)。通過(guò)Eviews軟件，得到上述模型的散點(diǎn)圖:X2X3 -X4 -X5X6YX1圖1散點(diǎn)圖2.2多元回歸對(duì)于多元回歸模型的建立，應(yīng)該滿足以下基本假設(shè)：解釋變量是非隨機(jī)的或固定的，且相互之間互不相關(guān)，即無(wú)多重共 線性。隨機(jī)干擾項(xiàng)具有零均值，同方差及不序列相關(guān)性，即E(7) =0,i =1,2,，n Var(7) = E(叫)*2,i =1,2,，nCov(7,j)二 E(7，U)=0,i = j,i, j =1,2，n解釋變量與隨機(jī)干擾項(xiàng)不相關(guān)，即Cov(Xji,7) =0,j =1,

11、2,，k,i =1,2，n隨機(jī)干擾項(xiàng)滿足正態(tài)分布叫 N(0,；2)樣本容量趨于無(wú)窮時(shí)，各解釋變量的方差趨于有界常數(shù)?；貧w模型的設(shè)定是正確的。再助于計(jì)量經(jīng)濟(jì)學(xué)軟件Eview3.1，對(duì)數(shù)據(jù)進(jìn)行最小二乘估計(jì)結(jié)果如圖所示:Depe ndent Variable: YMethod: Least SquaresDate: 06/01/12 Time: 20:40Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X1-0.0301830.023155-1.3035140.2190

12、X21.2390301.0836631.1433720.2772X31.0099080.3629652.7823840.0178X43.2316081.9401841.6656190.1240X5-1.3816850.864258-1.5986950.1382X60.2730280.1223992.2306310.0475C897.54541102.4220.8141570.4328R-squared0.994299Mean depe ndent var6922.547Adjusted R-squared0.991189S.D.dependent var5673.176S.E. of regr

13、essi on532.5272Akaike info criteri on15.67845Sum squared resid3119438.Schwarz criteri on16.02470Log likelihood-134.1060F-statistic319.7300Durb in -Watson stat1.950283Prob(F-statistic)0.000000圖2最小二乘估計(jì)的回歸結(jié)果估計(jì)模型結(jié)果如下：Yt =897.5454-0.030183X!t 1.239030X2( 1.009908乂氏 3.231608X4t -1.381685乂戢 0.273028X&0.814

14、157? ：i-1.3035141.1433722.7823841.665619-1.598695 ?2.2306312 2R =0.994299R =0.991189F =319.7300 S.E= 532.52723回歸模型的檢驗(yàn)和預(yù)測(cè)3.1經(jīng)濟(jì)意義上變量的檢驗(yàn)3.1.1擬合優(yōu)度檢驗(yàn)從Eviews回歸結(jié)果來(lái)看，模型擬合優(yōu)度很好。R2 =0.9942993.1.2對(duì)方程的T檢驗(yàn)1）對(duì)X1t進(jìn)行檢驗(yàn)提出原假設(shè)H0：Bj=O;備擇假設(shè) 出：片式0 , （ j電 ）|t |=1.303514假設(shè)顯著水平口 =0.05查自由度為11to.025（18-6-1）=2.201 1.303514，故接受

15、H。，拒絕 H）對(duì)X2t進(jìn)行檢驗(yàn)提出原假設(shè)H。： =0;備擇假設(shè)已：=0，（ j假設(shè)顯著水平a =0.05查自由度為11t.025（18-6-1）=2.201 =1.143372，故接受 H。，拒絕 H）對(duì)X3t進(jìn)行檢驗(yàn)提出原假設(shè)H。：打=0;備擇假設(shè) 比：=0，（ j假設(shè)顯著水平a =0.05查自由度為11to25（18-6-1）=2.201 2.782384,故拒絕原假設(shè) H。， 著的。4 ）對(duì)X4t進(jìn)行檢驗(yàn)提出原假設(shè)H0: ： j = 0;備擇假設(shè) 出：j = 0， （ j假設(shè)顯著水平a =0.05查自由度為11to.o25（18-6-1）=2.201 1.665619,故接受 H。，拒

16、絕 H5）對(duì)X5t進(jìn)行檢驗(yàn)提出原假設(shè)H。：打=0;備擇假設(shè)比：打=0，（ j假設(shè)顯著水平口 =0.05查自由度為11to.o25（18-6-1）=2.201 1.598695,故接受 H。，拒絕 H6）X6t進(jìn)行檢驗(yàn)提出原假設(shè)Ho: 1 =0;備擇假設(shè)H1j = 0，（ j假設(shè)顯著水平口 =0.05查自由度為11to25（1861）=2.201 2.230631,故拒絕原假設(shè) H。， 著的。的分布表，得臨界值 1，即變量X1t是不顯著的。=3）|t |=1.143372的分布表，得臨界值 !，即變量X2t是不顯著的。=2）|t |=2.782384的分布表，得臨界值接受H1，即變量Xgt是顯=

17、2）|t |=1.665619的分布表，得臨界值，即變量X4t是不顯著的。=2）|t |=1.598695的分布表，得臨界值I,即變量X5t是不顯著的。=2）|t |=2.230631的分布表，得臨界值接受H1，即變量X6t是顯3.1.3對(duì)方程的F檢驗(yàn)F =319.7300假設(shè)顯著水平:=0.05，查自由度為6和11的F分布表，得 臨界值Fo.o5（6,11） = 5.O7v319.73O0是顯然的，故F統(tǒng)計(jì)量的值在給定顯著性水平 下=0.05的情況下是顯著的。3.2計(jì)量經(jīng)濟(jì)學(xué)檢驗(yàn)3.2.1多重共線性檢驗(yàn) 計(jì)量經(jīng)濟(jì)學(xué)中多重共線性產(chǎn)生的原因：1）經(jīng)濟(jì)變量相關(guān)的共同趨勢(shì)；2）滯后變量的引入；3）樣

18、本資料的限制。 多重共線性的后果有：1）完全共線性下參數(shù)估計(jì)量不存在；2）近似共線性下普通最小二乘法參數(shù)估計(jì)量的方差變大；3）參數(shù)估計(jì)量經(jīng)濟(jì)含義不合理；4）變量的顯著性檢驗(yàn)和模型的預(yù)測(cè)功能失去意義?？朔嘀毓簿€性的方法：1）排除引起共線性的變量；2）差分法；3）減小參數(shù)估計(jì)量的方差。 多重共線性的檢驗(yàn)：1）檢驗(yàn)多重共線性是否存在；2）判明存在多重共線性的范圍（判定系數(shù)檢驗(yàn)法、逐步回歸法）。下列用相關(guān)系數(shù)檢驗(yàn)法檢驗(yàn)解釋變量的多重共線性，經(jīng)過(guò)計(jì)算得到變量 之間的相關(guān)系數(shù)如圖所示：Correlation MatrixX1X2X3X4X5 | X6X1X2X3X4X5|X&X1xi1.0000000.

19、9317630.9066290.8901150.6102590.052538X2X20.9317631 0000000 9028060 8111560.5288840.787882X3X30 9866290 9028061 0000000 8258620.5647830 896160X4X4Q8961150 3111560.0258621 0000000 &373420771061X5X50.6182590.5288840 5&47830.6373421.0000000.73&453X6X60.S925380.7078820.896160077106107394531.000000圖3相關(guān)系數(shù)

20、矩陣由上圖知，相關(guān)系數(shù)在0.90以上，這說(shuō)明解釋變量之間高度線性相關(guān)，即 存在比較嚴(yán)重的多重共線性，也是貨幣供應(yīng)量X1t與黃金儲(chǔ)備X2t與外匯儲(chǔ)備X3t之間存在比較嚴(yán)重的多重共線性。由于多重共線性的存在，我們采用逐步回歸法對(duì)模型進(jìn)行修正:第一步：運(yùn)用OLS方法逐一求Y對(duì)各個(gè)解釋變量的回歸：Depe ndent Variable: YMethod: Least SquaresDate: 06/01/11 Time: 12:41Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticP

21、rob.X10.0277570.00142619.457960.0000C226.6330442.09700.5126320.6152R-squared0.959454Mean depe ndent var6922.547Adjusted R-squared0.956920S.D.dependent var5673.176S.E. of regressi on1177.512Akaike info criteri on17.08463Sum squared resid22184563Schwarz criteri on17.18356Log likelihood-151.7617F-stati

22、stic378.6123Durb in -Watson stat0.445939Prob(F-statistic)0.000000圖4兀回歸1Depe ndent Variable: YMethod: Least SquaresDate: 06/01/11Time: 12:44Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X27.6611220.9543138.0278950.0000C-6707.8661805.727-3.7147730.0019R-squa

23、red0.801112Mean depe ndent var6922.547Adjusted R-squared0.788681S.D. dependent var5673.176S.E. of regressi on2607.928Akaike info criteri on18.67494Sum squared resid1.09E+08Schwarz criteri on18.77387Log likelihood-166.0744F-statistic64.44709Durb in -Watson stat1.097939Prob(F-statistic)0.000001圖5 一元回歸

24、2Depe ndent Variable: YMethod: Least SquaresDate: 06/01/11Time: 12:50Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X30.6394200.01799135.541580.0000C2313.649201.440511.485520.0000R-squared0.987492Mean depe ndent var6922.547Adjusted R-squared0.986710S.D.depe

25、ndent var5673.176S.E. of regressi on654.0047Akaike info criteri on15.90855Sum squared resid6843554.Schwarz criteri on16.00748Log likelihood-141.1769F-statistic1263.204Durb in -Watson stat0.563353Prob(F-statistic)0.000000圖6 -兀回歸3Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 21:26Samp

26、le: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X48.8941801.5049555.9099310.0000C-2850.6751825.260-1.5617910.1379R-squared0.685827Mean depe ndent var6922.547Adjusted R-squared0.666191S.D.dependent var5673.176S.E. of regressi on3277.747Akaike info criteri on19.13

27、214Sum squared resid1.72E+08Schwarz criteri on19.23107Log likelihood-170.1893F-statistic34.92728Durb in -Watson stat0.163517Prob(F-statistic)0.000022圖7 一元回歸4Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12 Time: 21:27Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Error

28、t-StatisticProb.X59.0563533.6031632.5134450.0230C3765.0641714.6892.1957720.0432R-squared0.283071Mean depe ndent var6922.547Adjusted R-squared0.238263S.D.dependent var5673.176S.E. of regressi on4951.410Akaike info criteri on19.95717Sum squared resid3.92E+08Schwarz criteri on20.05610Log likelihood-177

29、.6145F-statistic6.317406Durb in -Watson stat0.527980Prob(F-statistic)0.023042圖8 -兀回歸5Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 21:28Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X61.5534630.1879948.2633800.0000C2637.633793.43363.324328

30、0.0043R-squared0.810164Mean depe ndent var6922.547Adjusted R-squared0.798300S.D.dependent var5673.176S.E. of regressi on2547.883Akaike info criteri on18.62835Sum squared resid1.04E+08Schwarz criteri on18.72728Log likelihood-165.6552F-statistic68.28345Durb in -Watson stat1.520117Prob(F-statistic)0.00

31、0000圖9一元回歸6第二步：對(duì)比分析，根據(jù)調(diào)整后的可決系數(shù)R2最大的原則，選取X3t作為進(jìn)入回歸模型的第一個(gè)解釋變量，形成一元回歸。再將其余解釋變量分別引入模型， 得到二元回歸模型如下：Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 21:48Sample: 1993 2010In eluded observati ons: 18VariableCoeffieie ntStd. Errort-StatistieProb.X30.6614550.1138585.8094590.0000X1-0.0009840.005

32、014-0.1961510.8471C2392.089450.65265.3080540.0001R-squared0.987524Mean depe ndent var6922.547Adjusted R-squared0.985861S.D.dependent var5673.176S.E. of regressi on674.5885Akaike info eriteri on16.01709Sum squared resid6826045.Sehwarz eriteri on16.16549Log likelihood-141.1539F-statistie593.6649Durb i

33、n -Watson stat0.556224Prob(F-statistie)0.000000圖10二元回歸1Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 21:49Sample: 1993 2010In eluded observati ons: 18VariableCoeffieie ntStd. Errort-StatistieProb.X30.6459970.04316514.965650.0000X2-0.0968980.574197-0.1687540.8682C2438.645769.31033.16

34、99110.0063R-squared0.987516Mean depe ndent var6922.547Adjusted R-squared0.985851S.D.dependent var5673.176S.E. of regressi on674.8128Akaike info eriteri on16.01776Sum squared resid6830585.Sehwarz eriteri on16.16615Log likelihood-141.1598F-statistie593.2652Durb in -Watson stat0.556326Prob(F-statistie)

35、0.000000圖11二元回歸2Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 21:49Sample: 1993 2010In eluded observati ons: 18VariableCoeffieie ntStd. Errort-StatistieProb.X30.6269430.03272019.160600.0000X40.2521630.5461330.4617240.6509C2126.498454.94084.6742300.0003R-squared0.987667Mean depe nden

36、t var6922.547Adjusted R-squared0.986023S.D.dependent var5673.176S.E. of regressi on670.7037Akaike info eriteri on16.00554Sum squared resid6747652.Schwarz eriteri on16.15394Log likelihood-141.0499F-statistie600.6491Durb in -Watson stat0.565155Prob(F-statistie)0.000000圖12二元回歸3Depe ndent Variable: YMet

37、hod: Least SquaresDate: 06/03/12Time: 21:50Sample: 1993 2010In eluded observati ons: 18VariableCoeffieie ntStd. Errort-StatistieProb.X30.6550000.02135930.665750.0000X5-0.7297490.565033-1.2915150.2161C2455.775225.969210.867740.0000R-squared0.988744Mean depe ndent var6922.547Adjusted R-squared0.987243

38、S.D.dependent var5673.176S.E. of regressi on640.7653Akaike info eriteri on15.91422Sum squared resid6158702.Sehwarz eriteri on16.06261Log likelihood-140.2279F-statistie658.8057Durb in -Watson stat0.683607Prob(F-statistie)0.000000圖13二元回歸4Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 2

39、1:50Sample: 1993 2010In eluded observati ons: 18VariableCoeffieie ntStd. Errort-StatistieProb.X30.6114410.04109114.880290.0000X60.0837410.1102140.7598050.4591C2284.336207.768710.994610.0000R-squared0.987956Mean depe ndent var6922.547Adjusted R-squared0.986350S.D. dependent var5673.176S.E. of regress

40、i on662.8186Akaike info criteri on15.98189Sum squared resid6589928.Schwarz criteri on16.13029Log likelihood-140.8370F-statistic615.2046Durb in -Watson stat0.492248Prob(F-statistic)0.000000圖14二元回歸5第三步：再根據(jù)調(diào)整后的可決系數(shù)R2最大原則和參數(shù)顯著性原則，選取X5t 作為進(jìn)入回歸模型的第二個(gè)解釋變量，形成二元回歸；根據(jù)上面選取解釋變量的原則，繼續(xù)進(jìn)行逐步回歸，使得調(diào)整后的可決系數(shù)R2最大和參數(shù)都顯著，

41、如圖15和圖16,依次引入解釋變量X6t和X4t，最后引入的解釋變量X1t和X2t，雖然 也使調(diào)整后的可決系數(shù)有所增加，但是它們的參數(shù)都不顯著，如圖17和圖18,故舍去X1t和X2t，所以最后得到的回歸模型是圖16。Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12 Time: 22:13Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X30.5794040.03665815.805490.0000X

42、5-1.7047920.639190-2.6671130.0184X60.2885670.1205312.3941360.0312C2544.661200.487712.692360.0000R-squared0.992014Mean depe ndent var6922.547Adjusted R-squared0.990302S.D. dependent var5673.176S.E. of regressi on558.6763Akaike info criteri on15.68215Sum squared resid4369669.Schwarz criteri on15.88001

43、Log likelihood-137.1393F-statistic579.6653Durb in -Watson stat0.743211Prob(F-statistic)0.000000圖15引入X6t回歸Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 22:18Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X30.5398840.04191812.879540.0000X5-2.

44、1129180.650296-3.2491660.0063X60.3179940.1149382.7666640.0160X40.7761050.4664001.6640330.1200C2037.833358.40485.6858440.0001R-squared0.993416Mean depe ndent var6922.547Adjusted R-squared0.991390S.D. dependent var5673.176S.E. of regressi on526.4076Akaike info criteri on15.60016Sum squared resid360236

45、4.Schwarz criteri on15.84749Log likelihood-135.4015F-statistic490.3750Durb in -Watson stat1.176105Prob(F-statistic)0.000000圖16引入X4t回歸Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 22:19Sample: 1993 2010In cluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X30.65715

46、80.1936423.3936770.0053X41.3643731.0608571.2861040.2227X5-1.9356580.724783-2.6706740.0204X60.2984130.1219002.4480100.0307X1-0.0064110.010322-0.6210900.5462C2084.834374.90125.5610220.0001R-squared0.993621Mean depe ndent var6922.547Adjusted R-squared0.990963S.D. dependent var5673.176S.E. of regressi o

47、n539.3027Akaike info criteri on15.67963Sum squared resid3490169.Schwarz criteri on15.97642Log likelihood-135.1167F-statistic373.8411Durb in -Watson stat1.172971Prob(F-statistic)0.000000圖17引入X1t回歸Depe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 22:20Sample: 1993 2010In eluded observati

48、ons: 18VariableCoeffieie ntStd. Errort-StatistieProb.X30.5418280.0544129.9579440.0000X40.7835420.5010631.5637610.1438X5-2.1104720.677984-3.1128620.0090X60.3168760.1210662.6173760.0225X2-0.0293360.490696-0.0597850.9533C2070.071655.65413.1572610.0083R-squared0.993418Mean depe ndent var6922.547Adjusted

49、 R-squared0.990676S.D. dependent var5673.176S.E. of regressi on547.8208Akaike info eriteri on15.71098Sum squared resid3601292.Schwarz eriteri on16.00777Log likelihood-135.3988F-statistie362.2317Durb in -Watson stat1.175479Prob(F-statistie)0.000000圖18引入X2t回歸3.2.2異方差性檢驗(yàn)計(jì)量經(jīng)濟(jì)學(xué)中異方差性檢驗(yàn)的后果有：1）參數(shù)估計(jì)量非有效；2）變量

50、的顯著性檢驗(yàn)失去意義；3）模型的預(yù)測(cè)失效。異方差性檢驗(yàn)的方法有：1）圖示檢驗(yàn)法;300002000010000E2圖19解釋變量與殘差平方 e2的散點(diǎn)圖1200000-800000-600000-400000-200000800000-0600000400000200000 1A0-200000-4000009496980002080406-100000010ResidualActualFitted圖20殘差e2由圖3解釋變量與殘差平方e2的散點(diǎn)圖和圖4殘差e2的信息都表明可能不存在異方差。2）帕克（Park）檢驗(yàn)與戈里瑟（Gleiser ）檢驗(yàn);Depe ndent Variable: LO

51、G（E2）Method: Least SquaresDate: 06/03/12Time: 23:21Sample: 1993 2010In eluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.X3-5.01E-060.000138-0.0362770.9716X40.0004400.0015380.2862890.7792X50.0001560.0021440.0727840.9431X6-8.68E-050.000379-0.2290870.8224C10.973751.1817909.2857010.

52、0000R-squared0.014686Mean depe ndent var11.23634Adjusted R-squared-0.288487S.D.dependent var1.529145S.E. of regressi on1.735756Akaike info criteri on4.170897Sum squared resid39.16705Schwarz criteri on4.418222Log likelihood-32.53807F-statistic0.048442Durb in -Watson stat2.732721Prob(F-statistic)0.995

53、019圖21 Park 檢驗(yàn)給定顯著水平a =0.05查自由度為13的分布表，得臨界值 t.025(13) = 2.160|t|，所以不存在異方差。3)G-Q (Goldfeld-Quandt )檢驗(yàn);G-Q檢驗(yàn)描述：將n =18組數(shù)據(jù)觀察值按可能引起異方差的解釋變量X1t的觀察值進(jìn)行升序排序；并將序列中間的c = n、4個(gè)觀察值除去，并將剩下的觀察值4劃分為較小與較大的容量相同的兩個(gè)子樣本，每個(gè)子樣樣本容量均為口 = 7 ；4對(duì)每個(gè)子樣分別進(jìn)行OLS回歸，并計(jì)算各自的殘差平方和RSS和RSS?；貧w結(jié)果如下：Depe ndent Variable: YMethod: Least Squares

54、Date: 06/03/12 Time: 23:37Sample: 1993 1999In eluded observati ons: 7VariableCoefficie ntStd. Errort-StatisticProb.X60.1212410.2643580.4586260.6915X5-1.2946150.980986-1.3197080.3177X4-1.5312240.356560-4.2944300.0502X32.4535430.11857420.692110.0023C1503.48445.2998433.189610.0009R-squared0.999785Mean

55、depe ndent var3005.095Adjusted R-squared0.999354S.D.dependent var812.5859S.E. of regressi on20.65808Akaike info criteri on9.069898Sum squared resid853.5123Schwarz criteri on9.031263Log likelihood-26.74464F-statistic2320.366Durb in -Watson stat2.737001Prob(F-statistic)0.000431圖 22 1993-1999年OLS回歸結(jié)果De

56、pe ndent Variable: YMethod: Least SquaresDate: 06/03/12Time: 23:39Sample: 2004 2010In eluded observati ons: 7VariableCoefficie ntStd. Errort-StatisticProb.X60.2106030.0797682.6401900.1185X5-1.1856310.442025-2.6822750.1154X4-1.7521361.889538-0.9272830.4517X30.6954710.05015713.865960.0052C3392.3322207

57、.2601.5368970.2641R-squared0.999177Mean depe ndent var12218.26Adjusted R-squared0.997532S.D. dependent var5996.254S.E. of regressi on297.8897Akaike info criteri on14.40713Sum squared resid177476.5Schwarz criteri on14.36850Log likelihood-45.42496F-statistic607.2715Durb in -Watson stat3.192268Prob(F-s

58、tatistic)0.001645圖 23 2004-2010年OLS回歸結(jié)果計(jì)算樣本1的殘差平方和：RSS= 853.5123和樣本2的殘差平方和:RSS =177476.5給定顯著性水平:=0.05，確定F分布表中的相應(yīng)臨界值 (2,2)=99.00 ；在同方差性假設(shè)下計(jì)算F的統(tǒng)計(jì)量：F二RS故表明存在RSS -z ss異方差。1）懷特（White）檢驗(yàn);White Heteroskedasticity Test:F-statistic1.362653Probability0.325854Obs*R-squared9.859796Probability0.275005Test Equati

59、 on:Dependent Variable: RESIDEMethod: Least SquaresDate: 06/03/12Time: 23:26Sample: 1993 2010In eluded observati ons: 18VariableCoefficie ntStd. Errort-StatisticProb.C345085.5360519.00.9571910.3635X6-39.10161220.9114-0.1770010.8634X6A2-0.0011700.016110-0.0725990.9437X51878.2541046.1631.7953740.1062X

60、5A2-1.3908590.645411-2.1549980.0596X4-1493.2521228.708-1.2153030.2552X4A20.9340780.8267491.1298210.2878X331.9191077.585730.4114040.6904X3A2-0.0025470.001827-1.3940790.1968R-squared0.547766Mean depe ndent var200131.4Adjusted R-squared0.145781S.D.dependent var281807.4S.E. of regressi on260457.5Akaike