我们计量的课程论文，残差检验有异方差，但是加权最小二乘法后又出现自相关问题，用一阶的广义差分法也消除不了，请问应该怎么办呢？
这是怀特检验的结果
White Heteroskedasticity Test:   
   
F-statistic 31.28428     Probability  0.000000
Obs*R-squared 25.46261     Probability  0.002500
   
   
Test Equation:   
Dependent Variable: RESID^2   
Method: Least Squares   
Date: 08/03/09   Time: 07:07   
Sample: 1983 2009   
Included observations: 27   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
C -1324631. 4238313. -0.312537 0.7584
X6 274.5861 986.4711 0.278352 0.7841
X6^2 0.145826 0.020480 7.120255 0.0000
X6*X5 -8.902415 4.936821 -1.803269 0.0891
X6*X4 -0.333687 0.033927 -9.835300 0.0000
X5 71237.29 49820.57 1.429877 0.1709
X5^2 -174.6775 230.2971 -0.758488 0.4585
X5*X4 21.05025 3.586716 5.868949 0.0000
X4 -2281.910 846.3889 -2.696054 0.0153
X4^2 0.119567 0.013572 8.809634 0.0000
   
R-squared 0.943060     Mean dependent var  4987819.
Adjusted R-squared 0.912915     S.D. dependent var  13268959
S.E. of regression 3915694.     Akaike info criterion  33.47700
Sum squared resid 2.61E+14     Schwarz criterion  33.95694
Log likelihood -441.9395     F-statistic  31.28428
Durbin-Watson stat 3.490797     Prob(F-statistic)  0.000000
   
权数用了1/X4之后
White Heteroskedasticity Test:   
   
F-statistic 0.551032     Probability  0.817621
Obs*R-squared 6.097683     Probability  0.730103
   
   
Test Equation:   
Dependent Variable: STD_RESID^2   
Method: Least Squares   
Date: 08/03/09   Time: 07:09   
Sample: 1983 2009   
Included observations: 27   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
C 19601.38 163731.9 0.119716 0.9061
X6 31.83019 38.10873 0.835247 0.4152
X6^2 0.000328 0.000791 0.414533 0.6837
X6*X5 -0.089872 0.190716 -0.471236 0.6435
X6*X4 -0.000220 0.001311 -0.167970 0.8686
X5 757.1888 1924.637 0.393419 0.6989
X5^2 -2.351286 8.896692 -0.264288 0.7947
X5*X4 0.121924 0.138560 0.879941 0.3912
X4 -37.09296 32.69717 -1.134440 0.2724
X4^2 -0.000217 0.000524 -0.414574 0.6836
   
R-squared 0.225840     Mean dependent var  133025.5
Adjusted R-squared -0.184009     S.D. dependent var  139018.1
S.E. of regression 151268.7     Akaike info criterion  26.96962
Sum squared resid 3.89E+11     Schwarz criterion  27.44956
Log likelihood -354.0899     F-statistic  0.551032
Durbin-Watson stat 2.364152     Prob(F-statistic)  0.817621
   
但是这时的DW值变小了

Dependent Variable: Y   
Method: Least Squares   
Date: 08/03/09   Time: 07:11   
Sample: 1983 2009   
Included observations: 27   
Weighting series: 1/X4   
   
Variable Coefficient Std. Error t-Statistic Prob.  
   
X6 1.862733 0.022604 82.40787 0.0000
X5 18.03562 3.486012 5.173710 0.0000
X4 0.823781 0.089580 9.196043 0.0000
   
Weighted Statistics   
   
R-squared 0.996726     Mean dependent var  17169.65
Adjusted R-squared 0.996454     S.D. dependent var  8162.580
S.E. of regression 486.0930     Akaike info criterion  15.31512
Sum squared resid 5670875.     Schwarz criterion  15.45910
Log likelihood -203.7541     F-statistic  3653.721
Durbin-Watson stat 0.693058     Prob(F-statistic)  0.000000
   
Unweighted Statistics   
   
R-squared 0.999061     Mean dependent var  94740.78
Adjusted R-squared 0.998983     S.D. dependent var  96092.44
S.E. of regression 3064.712     Sum squared resid  2.25E+08
Durbin-Watson stat 1.283074   
   
请问该怎么办呢？本人只是初学者，那位大侠给予赐教，感激不尽啊。谢谢