My initial test case was this double lowerbound = gsl_cdf_beta_Qinv(0.95, 12093.0, 3602.0);
which appears to be fixed with the patch, but I since found a new case that fails: double lowerbound = gsl_cdf_beta_Qinv(0.95, 9656, 5038); -- Hans Ekbrand <hans.ekbr...@gmail.com>
