Explanation of the Forecast Combination Puzzle --- where can i find it?
----------- 关于夸夸而谈之词，希望叔叔不是在笑话侄子，哈哈哈。。。我随口而出的一句话不必介意。我意思，打败别的的需要证据，尤其是当对方宣称有证据的时候。
------------
另外，我将我的do we always need a filter的核心意思（介绍给教研组的时候）解释成：
两块表并不一定比一块表读的时间更准。

和你的话相结合，那就是复杂的东西真心的不一定更好。。。。
还是那句，复杂的内容一定是基于更多模型假设之上才采用的，那么这些模型假设的质量非常关键，直接决定有没有这些复杂的东西是不是worth！就好比，我又块表显示时间15:27:30，结果你又拿来一块表，显示时间15:38:00,两者综合考虑的时间可能还没有我那一块表更准。得看你的那块表的质量~~

555

http://blog.sciencenet.cn/blog-107667-830457.html

As it is widely accepted that only “good” forecasts should beincluded in a combination, strong differences in forecast error variancesbetween the individual forecasts are not to be expected. In such circumstances, it seems unlikely that a weighted average will outperform a simple averagecombination.
  
《Lilian M. de Menezes, Derek W.Bunn, James W. Taylor.Review of guidelines for the use of combined forecasts. European Journal of Operational Research, 2000, 120(1): 190-204.》
侄子天才！