The first obvious place is fac_to_ren_ratio, that has been terrible for quite some time, and finally identified as wrong.
Renormalization scale is only affecting the process side, so it does not enter at all DGLAP evolution.
We should replace all occurrences with the process scale $Q^2$, possibly after replacing all occurrences of Q2 as factorization scale, as described in #157
At the beginning, we had the problem on how to split the yadism and evolution settings in the database, but this is no longer a problem: the variations are only fact and ren, there is no further degree of freedom. Whether this affects evolution or processes depends on the scheme choice, and it is resolved by Pineko.
There is no need to split them in separate theory options, simply EKO and yadism (w/ generator friends) will potentially receive two different theories, because only one of them should implement fact scvar.
This is already an improvement on its own. E.g. since EKO does not implement ren scvar, it should never receive the option in the first place.
So, we will trim theory cards that we are passing around NNPDF/pineko#61
The first obvious place is
fac_to_ren_ratio, that has been terrible for quite some time, and finally identified as wrong.Renormalization scale is only affecting the process side, so it does not enter at all DGLAP evolution.
We should replace all occurrences with the process scale$Q^2$ , possibly after replacing all occurrences of
Q2as factorization scale, as described in #157At the beginning, we had the problem on how to split the yadism and evolution settings in the database, but this is no longer a problem: the variations are only fact and ren, there is no further degree of freedom. Whether this affects evolution or processes depends on the scheme choice, and it is resolved by Pineko.
There is no need to split them in separate theory options, simply EKO and yadism (w/ generator friends) will potentially receive two different theories, because only one of them should implement fact scvar.
This is already an improvement on its own. E.g. since EKO does not implement ren scvar, it should never receive the option in the first place.
So, we will trim theory cards that we are passing around NNPDF/pineko#61