Wrong error on number of events when using AsymptoticError and Extended flag in tutorial rf611_weightedfits.C

[thucking]

Dear Rooters,

I made a small modification of the example macro rf611_weightedfits.C by making the data generation and the fits extended. Doing this the error on the number of events is smaller then 0.1e-3, in case the AsymptoticError flag is used in conjunction with the Extended flag.

Here is a plot of what I get:

The third row shows the fitted number of events and the last one the corresponding error.

You can see the error on the fitted number of events with AsymptoticError flag enbabled in the bottom right corner. Obviously the error is very low (< 0.1e-3).
The error on the number of events with SumW2Error(true) has some outliers but that is not what I’m concerned about here.

Can anybody tell me if this is an issue with the formula itself so that it can not provide correct errors in case of an extended fit or if it could be an issue with the implementation?

Based on my previous experience I can already say thanks for your replies

If you want to reproduce the problem you can find the macro here:
rf611_weightedfits_modTPH.C (11.2 KB)

Possibly related issue

SumW2Error with Extended ML

Hello, I wanted to fit a weighted dataset and obtain the normalization factor and its error with SumW2Erorr(kTRUE). But I found somewhat strange behavior of the option. After some test with a simple gaussian, I am quite confused about this option. Here’s the code used for the tests. test was done on the ROOT v5.34.18 using namespace RooFit; void gaussianSumW2() { RooRealVar x("x","x",-10,10); RooRealVar mu("mu","mu",0,-1,1); RooRealVar sig("sig","sig",1,0,10); RooGaussian gauss("…

ROOT version

• 6.20.04 compiled with gcc 7.5

[etejedor]

@moneta or @StephanH could you give a hand with this one? Thanks!

[StephanH]

That’s interesting. @moneta, could this be because extended only has an effect on the NLL, but not on the function itself? Therefore, derivatives w.r.t. normalisation parameters probably don’t work.

[moneta]

Hi, Yes @stephanh , you are correct ! The derivative of the extended term is missing and needs to be included when computing the correction as shown in equation 18 of https://arxiv.org/pdf/1911.01303.pdf

I have opened a JIRA ticket for this, https://sft.its.cern.ch/jira/browse/ROOT-10827

Thank you for reporting this!
Best regards

Lorenzo

[thucking]

Thanks a lot for solving the problem so fast

[thucking]

@moneta,

could you give a time estimate how long it takes until it is implemented?

Thanks a lot.

[moneta]

Hi,
It should be implemented in few days, hope to have a PR fixing this in the next days
Best regards

Lorenzo

[thucking]

Thanks a lot. I am looking forward to it and sorry for disturbing you.