FFT convolution shape depends on variable range

Roofit and RooStats
1

Hi Root experts,

I’m trying to obtain a convolution using a Breit Wigner and crystal ball resolution. However, I noticed that the peak shifts as my variable range changes. Please see the attached script and plot that illustrates the problem, plotting 4 convolutions with the same parameters and different variable ranges. I don’t think this effect is mentioned in the documentation. Can you comment?

Thanks!
Shanti
BWxCB.pdf (4.86 KB)
roofit_bwcb_fft2.C (2.19 KB)

2

Hi,

Sorry, after updating to v5.24.03, the problem is fixed.

Thanks!
Shanti

3

I just wanted to confirm that the bug Shanti discovered was fixed in v 5.24.03 and this FFT-based convolution technique is working sooooo beautifully!!! Just think about what you can do with this! - fitting for the numbers of events AND for the parameters of your detector resolution function AND for the underlying distribution’s parameters simultaneously now works really fast. That’s a sales pitch, yes :slight_smile: