How to choose the best hypothesis for testing two sample sets of one dimension each without considering distribution

I have tried run kstest to compare two data sets whose distribution is not known exactly. The data set is single dimension and i tried with kstest2(x1,x2). But it is not giving me the correct results. Kindly let me know the selection criteria for choosing the test measure and what should i implement it?

[the cyclist]

You need to provide more details, ideally including your code and a small sample of data. Why do you think the results of kstest2() are incorrect?

[Shashi]

Actually...it was a mistake on my part as i was trying to know the amount by which one value is larger than other. Thanks however for your comment.

Regards

Shashi