为什么在面板数据模型中加入距离变量后所有的固定效应都不可以做  一坐固定效应就出现near singular matrix  网上说是多重共线性
而去掉距离变量固定效应模型可以做  当然加入距离变量随机效应和混合模型还是可以做的，距离变量时两地间的距离，是非实时变量，其他变量每年都是不一样的。
Dependent Variable: LEX?   
Method: Pooled Least Squares   
Date: 02/26/10   Time: 12:43   
Sample: 1996 2007   
Included observations: 12   
Cross-sections included: 32   
Total pool (unbalanced) observations: 364   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
LGDP? 1.357462 0.073127 18.56316 0.0000
LYYCZ -1.900830 0.222216 -8.553975 0.0000
LPOP? -0.356183 0.080583 -4.420076 0.0000
LDIS? -1.484503 0.117100 -12.67720 0.0000
D1? 1.796384 0.185167 9.701406 0.0000
   
R-squared 0.608922     Mean dependent var  2.225597
Adjusted R-squared 0.604565     S.D. dependent var  2.480347
S.E. of regression 1.559732     Akaike info criterion  3.740546
Sum squared resid 873.3623     Schwarz criterion  3.794079
Log likelihood -675.7794     Hannan-Quinn criter.  3.761823
Durbin-Watson stat 0.566239   
   
Dependent Variable: LEX?   
Method: Pooled EGLS (Cross-section random effects)   
Date: 02/26/10   Time: 12:43   
Sample: 1996 2007   
Included observations: 12   
Cross-sections included: 32   
Total pool (unbalanced) observations: 364   
Swamy and Arora estimator of component variances   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
C -37.17221 4.498887 -8.262534 0.0000
LGDP? 1.461493 0.120570 12.12155 0.0000
LYYCZ 2.403414 0.442128 5.436013 0.0000
LPOP? -0.310095 0.142446 -2.176923 0.0301
LDIS? -1.496250 0.212910 -7.027607 0.0000
D1? 0.115218 0.202255 0.569669 0.5693
Random Effects (Cross)   
JAPAN--C -0.608722   
USA--C 0.856706   
REPKOREA--C -0.364327   
GERMANY--C 0.555673   
UK--C 0.130650   
SPAIN--C 1.233330   
CANADA--C 1.229468   
NERTHERLANDS--C 0.857544   
FRANCE--C -0.005394   
BELGIUM--C 0.806858   
RUSSIAN--C 0.109330   
POLAND--C 0.986942   
PORTUGAL--C 1.132892   
THAILAND--C 0.236296   
MALAYSIA--C 0.827124   
PHILIPPINES--C -0.009824   
ITALY--C -1.264803   
SWEDEN--C 0.037794   
DENMARK--C -0.020870   
AUSTRALIA--C -0.298508   
BRAZIL--C -4.109067   
UKRAINE--C 0.096545   
INDONESIA--C -0.283313   
ISRAEL--C -0.791934   
LITHUANIA--C 1.071630   
EGYPT--C -1.867201   
SINGAPORE--C 0.445900   
NEWZEALAND--C 0.343785   
GREECE--C -0.040678   
NORWAY--C -0.916201   
MEXICO--C -0.404626   
VIETNAM--C 0.026999   
   
Effects Specification   
   S.D.   Rho  
   
Cross-section random   0.769726 0.3540
Idiosyncratic random   1.039721 0.6460
   
Weighted Statistics   
   
R-squared 0.548124     Mean dependent var  0.804416
Adjusted R-squared 0.541813     S.D. dependent var  1.620652
S.E. of regression 1.104849     Sum squared resid  437.0076
F-statistic 86.85046     Durbin-Watson stat  0.786043
Prob(F-statistic) 0.000000   
   
Unweighted Statistics   
   
R-squared 0.649157     Mean dependent var  2.225597
Sum squared resid 783.5099     Durbin-Watson stat  0.438421
   
而本身我的这个建立固定效应模型应该是最好的，因为选择的样本占全部的85%以上，不明白为何距离变量ldis?加入后就不可以做固定效应分析了