Why is it taking so long to do a Spearman correlation using Prism (or InStat)?

This problem occurred with Prism versions up to version 5.  With large samples (13 or more pairs), calculating the exact P value would take a very long time*, so Prism computes an approximate P value using a Gaussian assumption. With large sample sizes, this approximation works very well. The problem is with intermediate numbers of pairs (10-12). In that range, the exact computation is slow, and the approximation isn't super accurate. So Prism computes the exact P value, which can take many minutes. You can click "Cancel" to interrupt the computation and get the approximate value. That may be all you need.

If you interrupt the calculations, you will view approximate answers as you expect. The next time you choose Spearman correlation (in the same Prism session, but it could be a new analysis or an update of an existing one), Prism 5 will again report approximate results (no exact calculations will be done) but Prism incorrectly reports that the calculation type is exact (not approximate).

Prism 6 does these caculations much faster than did Prism 5, orders of magnitude faster, so you never have to wait long for the results.

______

*The exact calculations figure out every possible permutation of your pairs. Thus, 10 pairs take 10 times as long as 9, 11 pairs take 11 times as long as 10...