Dependent Variable: Y    
Method: ML - ARCH (Marquardt) - Normal distribution    
Date: 05/17/09   Time: 11:42    
Sample: 1 296    
Included observations: 296    
Estimation settings: tol= 0.00010, derivs=accurate mixed (linear)    
Initial Values: C(1)=2.94017, C(2)=0.62576, C(3)=0.00014, C(4)=0.17143    
Convergence achieved after 32 iterations    
Presample variance: backcast (parameter = 0.7)    
GARCH = C(3) + C(4)*RESID(-1)^2    

c(1),c(2)是均值方程的参数，直接对原数据用OLS可以得到。

但是c(3),c(4)不知道软件是怎么设定？可以肯定的是：eviews不是用ls  e^2 c e(-1)^2的结果作为c(3),c(4)的初始值。