Why are some values blank when I ask Prism to create a nonparametric correlation matrix?

 Prism 5 can create a correlation matrix showing the correlation coefficient between every column and every other column. You can choose to compute an ordinary or nonparametric (Spearman) correlation coefficient.

A bug can cause Prism (up to 5.02 and 5.0b) to leave parts of this matrix blank. This happens when it encounters a perfect negative correlation, so rs = -1. That triggers a math error, and Prism can't compute the rest of the results. We have fixed this in 5.03 and 5.0c. 

To bypass the problem, leave one of the columns out of the analysis, or compute the correlation coefficients a pair of columns at a time, each in its own analysis.