ROOT::FIT, parameter error propagation

ROOT
1

Dear rooters,

I’m still new to the ROOT::FIT classes. In some of the fits, several parameters are kept fix; but they have an associated error.

In order to propagate these errors (associated with the fix parameters) to the “free” parameters, I have to perform several fits changing the values of the fixed parameters (a Montecarlo way of propagating errors).

So, my question is: is there a way using ROOT::FIT to propagate the errors of the fixed parameters?

I spent a while looking for it, but I didn’t find anything. So, in case of a negative answer, is there any more stylish way of propagating the errors? (I can provide a “simplified working code” if needed).

Thank you for your time.

Greetings.

2

Hi,

The correct way to include the uncertainty on those parameter is to include an additional term in the likelihood, such as a Gaussian constraint.
This is not foreseen actually in ROOT::Fit, but it is not complicated to do it by extending the negative log-likelihood with a quadratic term in the parameter (the log of a gaussian) and then use the ROOT::Fitter::FitFCN function to perform the fit.
If you need I could make an example for you

Best Regards

Lorenzo

3

Hello

I would be grateful if you could make an example of the propagation of systematic errors.

I have simplified an example, if you don’t want to make it from scratch.
propagateSystematicErrors.C (2.2 KB)

Thank you for your time :slight_smile:

Best regards,
Alvaro.

4

Hi,

Here is an example where the constraint is added for the parameter 1, slope of the exponential

LorenzopropagateSystematicErrors.C (2.4 KB)

1 Like
5

Hello

Thank you for your example code.

The final error is smaller using this way, compared to the “Montecarlo way”.

I will need more time to understand the statistics that is causing this issue (do you have any suggestion?).

Thank you again for your time.

Best regards,
Alvaro.