and it is better to start from the highest possible lag, and reduce the lag each time.

because starting from only one lag and adding extra lag each time may lead to inappropriate inference.

你不是使用时间序列的资料吗？！

the | possible | highest means whatever you think reasonable by economic restriction, intuition.

for Quarterly data, it is then intuitive to start from 4 or 8. Sometimes people use the rule of Thumb: T to the power of 1/3, where T is the sample size.

我也在学那方面的知识

我觉得应该是通过AIC ,BIC来具体判断

不是你觉得而已， 一般来说，以三楼说的方式为通用！

这个问题，怎麽会有有这麽多的疑惑？！

Lag的选择，难道是在VAR、VECM这边才第一次碰到的吗？

在基础计量的 一些统计检定中，都已经碰过了，

重点就在为什麽要放Lags，如果还不懂，拜托，回去好好地把你基础计量的书拿出来再读几次！

根据AIC SIC 准则选择VAR的滞后阶数   然后ECM是 VAR的 减1吧