The entire covariance matrix has been scaled. Instead, one should extract uncertainties, $\sigmai = \sqrt{{\rm diag}(\Sigma{ij})}$ and correlation matrix, $\rho$ and then only scale $\sigma_i$ to achieve correct covariance matrix.
To Reproduce
run PAD/PADforSFS with any luminosity scaling setting.
System Settings
All systems
Describe the bug
During the rescaling of the covariance matrix in
https://github.com/MadAnalysis/madanalysis5/blob/774ebff6f478840e41ce2ff80398f807380f4ece/madanalysis/misc/run_recast.py#L1093-L1109
The entire covariance matrix has been scaled. Instead, one should extract uncertainties, $\sigmai = \sqrt{{\rm diag}(\Sigma{ij})}$ and correlation matrix, $\rho$ and then only scale $\sigma_i$ to achieve correct covariance matrix.
To Reproduce
run PAD/PADforSFS with any luminosity scaling setting.
Expected behaviour
No response
Log files
No response
Additional information
Thanks to @mdgoodsell for pointing this out!