ctools

To calculate spectral points or analyse weak sources, it is necessary to compute upper limits to properly handle the lack of statistics. Since ctlike provides full access to the likelihood function, we can calculate upper limits for any parameter for any source in the xml model.
In general, the upper limit value can be derived by inspecting the likelihood function and find the parameter value at which the likelihood changes by a certain value. For 95% confidence level, this value is, e.g., 1.92 (see Minuit manual)

The question remains, wether this should be implemented as a ctlike option, which could also add the UL value to the final xml model, or if we want to have a separate tool for that.

I attach a first approach using scipy to compute the upper limit value in a cs-script.