Dear Professor Andrew, 

     Thanks for your suggestions! I have got consistent results by using permutation number of 100,000 and 110,000.

Next step, I will try NBS but not sure how to set the primary statistic threshold.

The manual told that I can find the appropriate threshold by trail and error.

But the problem is, I have no reference or gold stand to decided which results is reasonable or not.

Thanks for your help again!

Hank