The PoI range when generate the workspace

Roofit and RooStats
1

Dear experts,

When we generate the workspace, we normally give the POI a range like [-50,50]. Then later when we do HypoTestInverter scanning, we need to give another range for the scanning like [-10,10].
I tested the following two cases:
A: workspace range [-50,50], scanning range [-10,10]
B: workspace range [-100,100], scanning range [-10,10]

I notice A could give out more reasonable results, e.g. less failed fitting in A case. But I naively think A and B should always give out the same results, since the scanning range is always smaller than the workspace range, hence for both cases the final used ranges for POI would be essentially the same. Could you explain why the workspace range can play role(s) here ?

Many Thanks!
Javier

2

Hi,

The workspace range can plays a role because it controls the region where the parameter is allowed to vary in the fit. In particular if the range is changed that allows to include or not the global minimum of the negative log-likelihood then is having a big influence.
The scanning range drives only the regions where the poi values are tested

Cheers

Lorenzo

3

Dear Lorenzo,

Many thanks for the reply! I see now.

Best Regards,
Javier