Bug: "How effective was the pairing" conclusion is backwards in early releases of Prism 6

As part of the paired t test calculations, Prism asks: How effective was the pairing? To find out, it computes a correlation coefficient, and the corresponding one-tailed P value. For the Wilcoxon matched pairs test, it computes the Spearman nonparameric correlation coefficient, also with a one-tailed P value. These calculations are correct. If your paired design worked, you expect significant correlation so the P value should be small. If the P value is large, you should question your experimental design as the pairing or matching was not effective.

Prism summarizes the results by asking: "Significant correlation? (P>0.05)". It then gives the answer, yes or no, to that exact question: Is the P value greater than 0.05. But, of course, it should be asking if the P value is less than 0.05. So the Yes/No conclusion Prism reports is confusing. Prism asks two questions: Is the correlation signficant, and is the P value greater than 0.05? Those two questions are inconsistent. When the answer to one is Yes, the answer to the other has to be No (and vice versa). Prism answers the second question correctly (is the P greater than 0.05), so Prism always answers the first question (is the correlation significant?) incorrectly.

This bug is in 6.00, 6.01, 6.0a and 6.0b. It was fixed in 6.02 (Windows) and 6.0c (Mac).