计量虾米向大家求助呀：
用eviews5.0做计量模型，DW值是1.016260，我查了表，临界值Dl和Du分别是0.90和1.83，所以不能确定是不是存在自相关。在估计方程中加入AR(1)，DW反而变成0.784723，再加AR(2)，DW是2.130257，但是解释变量又不显著了。
请问这种情况要怎么解决呀？附上计量结果，最上面是最初的结果，中间加AR(1)，最后加了AR(2).




Dependent Variable: LOG(EC)   
Method: Least Squares   
Date: 11/02/10   Time: 13:51   
Sample (adjusted): 1992 2009   
Included observations: 18 after adjustments   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
C 7.539466 0.003825 1971.256 0.0000
LOG(II) 0.008023 0.000664 12.08376 0.0000
LOG(IG) 0.001229 0.000578 2.125652 0.0533
LOG(IJ) -0.002939 0.001024 -2.869744 0.0131
LOG(PP) 0.000544 0.000178 3.054807 0.0092
   
R-squared 0.991545     Mean dependent var  7.600149
Adjusted R-squared 0.988944     S.D. dependent var  0.002671
S.E. of regression 0.000281     Akaike info criterion  -13.28717
Sum squared resid 1.03E-06     Schwarz criterion  -13.03985
Log likelihood 124.5845     F-statistic  381.1600
Durbin-Watson stat 1.016260     Prob(F-statistic)  0.000000




加AR(1)
Dependent Variable: LOG(EC)   
Method: Least Squares   
Date: 11/02/10   Time: 13:50   
Sample (adjusted): 1993 2009   
Included observations: 17 after adjustments   
Convergence achieved after 53 iterations   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
C 7.648718 0.018004 424.8434 0.0000
LOG(II) 0.000570 0.000247 2.309166 0.0414
LOG(IG) -2.30E-05 4.16E-05 -0.553774 0.5908
LOG(IJ) -1.71E-05 6.30E-05 -0.270741 0.7916
LOG(PP) -9.47E-06 1.26E-05 -0.752316 0.4677
AR(1) 0.991152 0.002675 370.5491 0.0000
   
R-squared 0.999965     Mean dependent var  7.600399
Adjusted R-squared 0.999949     S.D. dependent var  0.002526
S.E. of regression 1.81E-05     Akaike info criterion  -18.73249
Sum squared resid 3.60E-09     Schwarz criterion  -18.43842
Log likelihood 165.2262     F-statistic  62438.27
Durbin-Watson stat 0.784723     Prob(F-statistic)  0.000000
   
Inverted AR Roots       .99   
   





加AR(2)
Dependent Variable: LOG(EC)   
Method: Least Squares   
Date: 11/02/10   Time: 13:47   
Sample (adjusted): 1994 2009   
Included observations: 16 after adjustments   
Convergence achieved after 41 iterations   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
C 7.603252 0.012133 626.6836 0.0000
LOG(II) 0.001691 0.000581 2.908978 0.0173
LOG(IG) 6.76E-05 5.74E-05 1.176818 0.2695
LOG(IJ) -0.000206 0.000107 -1.917099 0.0875
LOG(PP) -2.07E-05 1.90E-05 -1.093344 0.3026
AR(1) 1.283394 0.180426 7.113124 0.0001
AR(2) -0.300835 0.176864 -1.700942 0.1232
   
R-squared 0.999943     Mean dependent var  7.600650
Adjusted R-squared 0.999905     S.D. dependent var  0.002381
S.E. of regression 2.33E-05     Akaike info criterion  -18.19994
Sum squared resid 4.87E-09     Schwarz criterion  -17.86194
Log likelihood 152.5995     F-statistic  26194.05
Durbin-Watson stat 2.130257     Prob(F-statistic)  0.000000
   
Inverted AR Roots       .97           .31