Been a while since my last post! A result of my work taking over!
But since then I’ve discovered the IDL routine for a two-sample KS test with kuiper statistic is not very functional. This was tested against Pat’s iraf KS test, and against python’s kuiper statistic. (and its internal KS test).
found using python: the probability of obtaining these two curves (BAL vs. nonBAL) from the same distribution is 0.00004237 roughly one part in 100 000 or better than 6 sigma. Though the creator of the the kuiper statistic warned that the probability is an approximation, especially for small sets. granted this set is far from small…still worried at how much of an approximation it is.
NOTE: when doing a test of a distribution against itself…the code breaks…(as we joked about earlier)