I’m using the Feldman-Cousins method to extract confidence intervals;
I can use it with simple models, however when I go to a more realistic model I find the following problem:
when performing the Neyman construction it is unable to find a single point inside the interval.
The typical output is
NeymanConstruction: Prog: 53/100 total MC = 4 this test stat = 0.117056
N_sig=525 N_bkg=32589.4 [-1e+30, -1e+30] in interval = 0
where I think the problem is in the [-1e+30, -1e+30] interval of the test statistics.
So I believe that in this case the error lies in the InverseCDF of the sampling distribution. Am I right?
How can I verify (and solve) such a problem?