The Bonferroni inequality is a way to control the familywise Type I error rate to a reasonable level when you are making a large number of pairwise comparisons, or a large number of contrasts.
If you want to do 10 comparisons but don't want alpha to be more than .05, then divide .05 by the number of comparisons to get the comparisonwise error rate.
Here's a link to a good reference (scroll down to Bonferroni):
http://psych.rice.edu/online_stat/chapter10/specific_comparisons.html
In terms of matrix applications, there are literally
dozens of applications to statistics, including probability, solving systems of linear equations, determining whether an experimental design has enough resolution to estimate all the desired effects, etc.