input                 ///
i   t  y   x
1 1  1  2
1 2  3  4
2 1  5  6
2 2  7  8
3 1  4  5
3 2  4  7

end

set seed 2012

sort i t

reshape wide y x, i(i) j(t)

sample 80                //draw a 80% sample

reshape long y x, i(i) j(t)

also, there are a couple of other ways to handle this issue.

please see

http://www.stata.com/support/faq ... /sampling-clusters/

谢谢，我刚刚用sample实现了，你这个reshape不错

是， 这个更简单


bsample 80, cluster(i)