另外 接下来的LM检验和GARCH都用平方后的数列做么？根据自相关图是不是说明应该建立AR（1）然后进行LM检验，我检验了一下，lags的值选了1，2,3,4 ，结果貌似都挺好 实际上，这种情况下应该选lags为几呢？

You don't need to do the ADF test for the return squared.

Q-stat is basically the summation of the squared act up to k's order, so you can just look at the p-value on the right. The"prob" on you table. If p-value is large, this means the acf are not significantly from zero which means no heteroscedasity.

You need to do the diagnostic test. You can first pick GARCH(1,1) and than further try GARCH(1,2) GARCH(2,1) or higher order to see if there is any difference. Usually GARCH(1,1) is enough.