計(jì)量模型報(bào)告

1、一：經(jīng)濟(jì)計(jì)量模型的設(shè)立被解釋變量y=國(guó)內(nèi)旅游收入解釋變量x1=國(guó)內(nèi)旅游人數(shù)解釋變量x2=城鎮(zhèn)家庭收入建立回歸模型如下：Y = P +P X +P X +p01122其中為隨機(jī)擾動(dòng)項(xiàng)二：數(shù)據(jù)的收集年份國(guó)內(nèi)旅游收入3)國(guó)內(nèi)旅游人數(shù)(xl)城鎮(zhèn)家庭可支配收入(x2)199448108.55126.8817042.94199559810.56038.0420019.26199670142.56909.8222974.03199778060.88234.0413091.72199883024.39262.815369.3199988479.210682.5815947.76200098000.51258

2、1.5132917.732001108068.215301.3837213.492002119095.717636.4543499.91200313517420017.3155566.62004159586.724165.6870477.42005184088.628778.54887773.62006213131.734804.35109998.22007259258.945621.97137323.92008302853.454223.79172828.4資料來(lái)源中國(guó)統(tǒng)計(jì)年鑒三：模型的估計(jì)解釋變量X1與X2分別與被解釋變量Y的散點(diǎn)圖如下所示:Dependent Variable: Y Me

3、thod: Least SquaresDate: 12/12/10 Time: 17:33Sample: 1994 2008Included observations: 15Coefficient-StatisticProb.VariabletStd. ErrorC33056.201996.41016.557820.0000X15.0045350.08731657.315130.0000X20.0077220.0059391.3001990.2180R-squared0.996925Mean dependent var133792.2Adjusted R-squared0.996413S.D.

4、 dependent var75865.01S.E. of regression4543.771Akaike info criterion19.85776Sum squared resid2.48E+08Schwarz criterion19.99937Log likelihood-145.9332F-statistic1945.409Durbin-Watson stat1.003533Prob(F-statistic)0.000000有關(guān)回歸結(jié)果如下：Y = 33056.20 +5.004535*X1 + 0.007722*X2(16.55782)(57.31513)(1.300199)r

5、2= 0.996925R2 = 0.996413SE=4543.771F=1945.409DW=1.003533N=15由此可知，當(dāng)其他因素不變的情況下，國(guó)內(nèi)旅游人數(shù)每增加一個(gè)百分點(diǎn)，國(guó) 內(nèi)旅游收入就增加5.004535個(gè)百分點(diǎn)。當(dāng)其他因素不變的情況下，城鎮(zhèn)家庭 可支配收入每增加一個(gè)百分點(diǎn)國(guó)內(nèi)旅游收入就增加0.007722個(gè)百分點(diǎn)。模型擬合情況良好。但重要解釋變量X2沒(méi)有通過(guò)t檢驗(yàn)，模型可能存在多 重共線性問(wèn)題。四.多重共線性問(wèn)題的檢驗(yàn)與處理采用方差擴(kuò)大化因子檢驗(yàn)得到VIF=162.8678 10,模型存在嚴(yán)重的多重共線 性問(wèn)題。利用嶺估計(jì)對(duì)模型進(jìn)行估計(jì)。嶺跡圖如下:由嶺跡圖可以判斷，當(dāng)lmd

6、 = 0.8是圖形趨于平穩(wěn)，則取lmd = 0.8進(jìn)行嶺估計(jì)。其估計(jì)方程結(jié)果如下：Y = 39530.7749314 +4.586225*X1 + 0.017038*X2擬合優(yōu)度nh=0.017038比較小，則可以認(rèn)為嶺回歸比較成功的解決了多重共 線性問(wèn)題。五.異方差問(wèn)題的檢驗(yàn)與處理從如下殘差圖可以看出，模型殘差有明顯增大的趨勢(shì)，則可以說(shuō)模型可能存 在異方差問(wèn)題。320000280000240000200000160000120000800004000094 95 96 97 98 99 00 01 02 03 04 05 06 07 08Residual Actual Fitted下面采用等

7、級(jí)相關(guān)檢驗(yàn)來(lái)判斷假設(shè)是否正確。對(duì)解釋變量X1進(jìn)行等級(jí)相關(guān)檢驗(yàn)，RS=-0.207143帶入t=RS*(15-2)A0.5/(1-RSA2)A0.5=-0.7634,由此可知，假設(shè)不成立，同理對(duì)解釋變量X2進(jìn)行等級(jí)相關(guān)檢驗(yàn)，假設(shè)也不成立。則可以說(shuō)模型不存在異方差問(wèn)題。六自相關(guān)問(wèn)題的檢驗(yàn)與處理采用DW檢驗(yàn)判斷是否存在自相關(guān)問(wèn)題。有上述回歸報(bào)告知DW=1003533,查DW檢驗(yàn)表，k=2,n=15，得dl=0.95.du=1.54dlDWdu因而不能判斷是否存在自相關(guān)問(wèn)題。要利用連貫檢驗(yàn)來(lái)判斷是否存在自相關(guān)問(wèn)題。由上圖可知，N+=9, N-=6,LN=8查連貫檢驗(yàn)表，得：Na=4,Nb=13。即得N

8、a LN Nb,所以否定自相關(guān)問(wèn)題。七虛擬變量的處理解釋變量X2與被解釋變量Y的散點(diǎn)圖如下所示：從此散點(diǎn)圖可以明顯看出存在一個(gè)異常點(diǎn)，可以對(duì)異常點(diǎn)進(jìn)行處理。12005年S=0其余年份輸入命令ls y c x1 x2 S,得到回歸報(bào)告如下:Dependent Variable: Y Method: Least SquaresDate: 12/28/10 Time: 23:07Sample: 1 15Included observations: 15Coefficient-StatisticProb.VariabletStd. ErrorC30970.792595.80411.931090.000

9、0X15.7893470.6476838.9385540.0000X2-0.2321580.196317-1.1825680.2619S192612.8157564.41.2224390.2471R-squared0.997293Mean dependent var133792.2Adjusted R-squared0.996555S.D. dependent var75865.01S.E. of regression4452.973Akaike info criterion19.86371Sum squared resid2.18E+08Schwarz criterion20.05252Log likelihood-144.9778F-statistic1350.867Durbin-Watson stat1.554524Prob(F-statistic)0.000000y = 30970.79 + 5.789347*X1 - 0.232158*X2 + 192612.8*S(11.93109) ( 8.938554)(-1.182568)(1.222439)r 2= 0.997293R2 = 0.996555SE=4452.973DW=1.554524S變量未通過(guò)