liyyu 发表于 2005-4-28 11:23
用状态空间模型做似然估计时，Log likelihood为负是否正常（红色部分）？有什么含义？请指教
具体输出如下：
Sspace: SSS1   
Method: Maximum likelihood (Marquardt)   
Date: 04/21/05   Time: 00:15   
Sample: 1989 2004   
Included observations: 16   
Valid observations: 9   
Estimation settings: tol= 0.00010, derivs=accurate numeric   
Initial Values: C(1)=-0.70131, C(2)=-0.64202, C(3)=-0.27113, C(4)=   
        -0.54477, C(5)=93.3475, C(6)=224.718   
User prior mean: SVEC0   
User prior variance: SVAR0   
Convergence achieved after 41 iterations   
   
       Coefficient Std. Error z-Statistic Prob.  
   
C(1) -0.389615 0.400720 -0.972288 0.3309
C(2) -0.602883 0.204864 -2.942848 0.0033
C(3) -0.177088 0.370944 -0.477399 0.6331
C(4) -0.510974 0.090720 -5.632460 0.0000
C(5) 4.500000 6.868143 0.655199 0.5123
C(6) 30.17909 13.03786 2.314727 0.0206
   
       Final State Root MSE z-Statistic Prob.  
   
SV1 2.021019 0.699744 2.888224 0.0039
   
Log likelihood  -20.14359      Akaike info criterion  5.809687
Parameters 6                      Schwarz criterion  5.941171
Diffuse priors 0                    Hannan-Quinn criter.  5.525947

galilee 发表于 2010-11-18 06:34
log likelihood 应该是负的,如果假设是高斯分布的话.