Re: Getting alpha and beta values from RooStats::ProfileLikelihoodCalculator

Hello Thiago,

Yes, there are direct method from the HypoTestResult object returned from the ProfileLikelihoodCalculator::GetHypoTest. (see http://root.cern.ch/root/html/RooStats__ProfileLikelihoodCalculator.html ) You can get the p value for the null hypothesis (alpha) with HypoTestResult::GetNullPValue() and the corresponding significance in terms of Gaussian sigmas. For the alternate p value you could use HypoTestResult::GetAlternatePValue(). The asymptotic distribution of the profile likelihood ration under H0 follows a central chi-square distribution, while under H1 it follows a non central chi square distribution and this is not yet implemented in RooStats. Currently if you want to get the alternate p value you need to run toys and calibrate your profile likelihood ratio. This can be done by using the HybridCalculator class.

Best Regards

Lorenzo
On Sep30, 2010, at 11:43 AM, Thiago Tomei wrote:

> Hello
>
> I am using RooStats::ProfileLikelihoodCalculator to make a hypothesis
> test. I would like to know if there are direct methods to get the
> value of beta (probability of making type II error) and the critical
> region (or critical point, as I am working in a 1-parameter problem).
>
> Thanks in advance
> Thiago
> --
> Thiago R. F. P Tomei - http://cern.ch/Thiago.Tomei
> SPRACE - São Paulo Regional Analysis Center
> CERN - European Organization for Nuclear Research
>
Received on Sat Oct 02 2010 - 10:40:06 CEST