前两个问题在Ｈｅｌｐ文件里找到了：

“EViews may stop iterating even when convergence is not achieved. This can happen for two reasons. First, the number of iterations may have reached the prespecified upper bound. In this case, you should reset the maximum number of iterations to a larger number and try iterating until convergence is achieved.

Second, EViews may issue an error message indicating a "Failure to improve"after a number of iterations. This means that even though the parameters continue to change, EViews could not find a direction or step size that improves the objective function. This can happen when the objective function is ill-behaved; you should make certain that your model is identified. You might also try other starting values to see if you can approach the optimum from other directions.

Lastly, EViews may converge, but warn you that there is a singularity and that the coefficients are not unique. In this case, EViews will not report standard errors or t-statistics for the coefficient estimates.”

但是还留有两问：

要解决的话朝什么方向改进呢？

造成missing value的数据溢出是怎么一回事啊？

- Alpha * (Inverse of Hessian) * Gradient

Direction: (Inverse of Hessian) * Gradient
Step: Alpha

Hint: The optimization needs a non-singular Hessian

这是Optimization常见的基本问题，但处理方法很繁琐；请先叁考相关教科书附录