好像可以用Q检验吧

我最近也比较受这个问题困扰，目前了解到的情况是可以用试错法，AIC标准，SIC标准（所谓的赤池，施瓦茨信息准则）等等，不过有待于实证验证

恩，一般用AIC和SC准则，eviews软件可以自动选择

最简单也是最不准备的方法是根据ACF和PACF图来确定ARMA（p,q)
另外比较简单的方法也是很多学者采用的方法可以根据AIC,SIC, HQ, modified AIC, modified SIC, Modified HQ值来确定， 一般选择使得某一个判定标准最小的lag length. 用这种方法比较简单，可以让eviews自动选择。具体的操作在eviews的操作手册里面有。 但是这种方法也有其缺陷。
The third method is based on sequential tests, which can be found in the article of Campbell and Perron (1991), Pitfalls and opportunities: What macroeconomists should know about unit roots. Ng and Perron (1995) demonstrate that a too parismonious lag length may lead to large size distortions, while an over-parameterized lag length may impair the power of test. However, the size problem is in general overweights the latter problem. They also indicate that methods based on sequential tests have an merit over information criterion due to less size distortions and more comparable power.

就是试呗，写个循环语句，例如lag从1到10，通过estat ic 取得AIC值，比较一下，取AIC值最小的那个滞后期就行了