Inverted hypothesis test fit failure

Dear roostate experts,

I’m trying to make a plot of CL scan wrt poi. I start working with the StandardHypoTestInvDemo.C which is in the roostate tutorial. However, my fit could not converge at any point of the poi while it does give a converged fit result of upper limit when I set the CL to 95%. I’m thinking if there’s any fitting config that I can tune with to make the fit successful? One thing to notice is that there’s more than 500 NPs in my workspace, mostly are gamma NPs. Would the fit fails because of too many NPs?

Attached at the end of this email you can find my input parameters for the StandardHypoTestInvDemo function and the error message. My workspace and the full log file can be found at

Thank you for your help!


Input parameters:

const char * wsName = "combined",
const char * modelSBName = "ModelConfig",
const char * modelBName = "",
const char * dataName = "obsData",
int calculatorType = 2,
int testStatType = 3,
bool useCLs = true ,
int npoints = 5,
double poimin = 0,
double poimax = 1.2,
int ntoys=1000,
bool useNumberCounting = false,
const char * nuisPriorName = 0

Error message (point 0.3 for example):

[#0] PROGRESS:Eval – Running for SigXsecOverSM = 0.3
----> Doing a re-scan first
----> trying with strategy = 1
----> trying with improve
[#0] ERROR:Minization – AsymptoticCalculator: failure in fitting for qmu or qmuA 0.3 return a dummy result
[#0] PROGRESS:Eval – P values for SigXsecOverSM = 0.3
CLs = nan +/- 0
CLb = nan +/- 0
CLsplusb = nan +/- 0

[#0] WARNING:Eval – HypoTestInverter - Skip invalid result for point SigXsecOverSM = 0.3
[#1] INFO:ObjectHandling – RooWorkspace::saveSnaphot(combined) replacing previous snapshot with name ModelConfig__snapshot


since your model is rather complex with many parameters it is possible that some of the fit do not converge for some of the points. One would need to try the fits starting maybe with different initial parameter values.

I can look in more detail at your workspace, but I could do only this in a few days

Best Regards


Hi Lorenzo,

Thank you for looking into this! Would you tell me which parameters I can tune with, so that I can already play a bit with it?



I cannot answer like this. I would need to look at your model in detail.


This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.