Am 07.08.2016 um 11:44 schrieb Riccardo (Jack) Lucchetti:
On Sat, 6 Aug 2016, Sven Schreiber wrote:
> in relation to the feature request about cross-section dependence
> testing in panels I've encountered some performance bottleneck.
The "smpl" command is rather CPU-expensive. Here's a different solution
(which can be further optimised a little, but I'm sure you get the idea):
OK, thanks for the great suggestions. Indeed my own previous and faster
code had only N=140 sample restrictions, whereas the later version had
(N-1)/2 more, so almost 70-fold, which then could explain going from
"instantaneous" to several seconds.
It's a pity though, because my "elegant" version is much more readable