Dear experts, For a while, I have been trying to troubleshoot a bias I see in my fits with HistFactory (with ROOT 6.22.06). For the purpose of this post, I set up toys with a very simple model of a Gaussian signal (mean = 0, sigma = 1) over a uniform background, in two channels. The model for Chan…

Dear experts, Sorry for bothering you again, but have one of you had a chance to look at this? Since last posting, I have also seen bias in the following variations of the above: Single channel fit with a known flat background (i.e. no ShapeFactor) Using TRandom3 instead of FillRandom to generat…

Hi, Sorry for the late reply. I will look at it later today Cheers Lorenzo

Dear @avenkate , I see one problem with how you generate the Gaussian signal template, you are filling the template with TH1::FillRandom, so statistical fluctuations in the template creation can cause a bias. You could create a function to create the template such that it represents the expected nu…

Hi Jonas, Thanks so much for getting back to me. I am still processing your reply. But I wanted to clarify some things you said: As long as the mean and sigma of the pull distribution are consistent with 0 and 1 respectively, then we can conclude that there is no significant bias, even if the me…

Hi Arvind, yes you summarized that correctly! You can also think about your perceived “bias” from another angle. In your script, you have set genYield = 10000, which is used both as signal and background yield in channel 1. Since you don’t include signal in your data (null hypethesis), your fitted n…

Home | News | Documentation | Download | Privacy Notice

ROOT Forum

Bias in HistFactory fit

Roofit and RooStats

show post in topic

Home
Categories
Guidelines
Terms of Service
Privacy Policy

Powered by Discourse, best viewed with JavaScript enabled