micro@ Obviously it depends on the underlying distribution. If it is already normal, n=1 is sufficient. In general, it's governed by the Berry–Esseen bound.
