Generate toy data using PDF

Roofit and RooStats
1

Dear Expert,

I have a 2D binned fit with different components. Now I want to test the fit using toy data generated directly from the PDFs.
I know that we can generate the toy data using PDF>generate(), but here I have ~5 PDfs of both variables (x and y).

Can you please help to do this?

Here is the script that I am using for 2D fit:
fit_2d.root (51.3 KB)
fit_2d_binned_final.C (13.1 KB)

Regards
Chanchal

2

Hi @chanchal!

I see in the script you already have your final RooAddPdf where you combine all the PDFs.

You can also use RooAbsPdf::generate() to create a toy dataset with two observables. Or maybe I’m missing the reason why you’re not just doing this?

std::unique_ptr<RooDataSet> toydata{model.generate({mD0, mD0pi})};

By the way, don’t forget to use the Extended() command argument if you also want to randomize the number of events in the toy dataset according to a Poisson distribution around the nominal sum of yields in the model.

Cheers,
Jonas

3

Hii @jonas,

Thanks for the reply. yes, I did this, and working fine.
Now I want to check the distribution of (fitted yield - input yield)/fit uncertainty in each toy dataset.

As far as I know in RooMCStudy, we can do this using mcstudy->plotPull().
Is there any way to do this here?

Here is the script:
fit_2d_toy_data.C (18.1 KB)

Regards
Chanchal

4

Hello!

Is this follow-up question in your last post in this thread still relevant, or should we just follow up here?