nnpdf icon indicating copy to clipboard operation
nnpdf copied to clipboard

Theory covmat for higher twist

Open RoyStegeman opened this issue 2 years ago • 6 comments

We want to add a covmat to account for higher twist effects. Therefore the covmat should be constructed similar to a 3pt prescirption where $\Delta^\pm = \pm \sigma_\mathrm{obs}\frac{c}{Q^2(1-x)}$ for come coefficient $c$.

impact on the datapoints is as follows: https://vp.nnpdf.science/xee4YTRKSAmRmN3iUZfkIg==

RoyStegeman avatar Nov 24 '23 17:11 RoyStegeman

Just a quick comment (also just as a reminder to myself) - we should allow for both theory covmat through scalevar and HT covmat to be added simultaneously.

RoyStegeman avatar Nov 24 '23 18:11 RoyStegeman

A comment so I don't forget, I think before doing this we should remove all the thcovmat stuff thst is not being used. That would facilitate adding new components (because one would know where to look...).

scarlehoff avatar Nov 25 '23 10:11 scarlehoff

Since I don't expect this PR to be one of the ones that's open for an infinite amount of time I think that would imply removing that functionality now-ish. I don't mind, but I had understood you wanted to keep it as it currently is to make sure nobody is using it (especially make_scale_var_covmat and down from there)

RoyStegeman avatar Nov 25 '23 10:11 RoyStegeman

No, I put the exception to make sure nobody was using it but I'd rather remove it as soon as possible.

scarlehoff avatar Nov 25 '23 11:11 scarlehoff

I had a look at the fitting part, and, it looks like it should work now even though it is "by accident". This is because the covmat is saved under the same name as a scalevar covmat and for the fit they will be treated in the same way. While the HT covmat implementation is living in this branch that is not a problem so with the pipeline set up all that's left to do before performing a fit is to make sure the computation of the covmat is done correctly.

RoyStegeman avatar Dec 11 '23 11:12 RoyStegeman

Greetings from your nice fit :robot: ! I have good news for you, I just finished my tasks:

  • Fit Name: NNBOT-51f3b1fd8-2023-12-18
  • Fit Report: https://vp.nnpdf.science/Pi29O_9oQEqpTHVRXm3ZIg==
  • Fit Data: https://data.nnpdf.science/fits/NNBOT-51f3b1fd8-2023-12-18.tar.gz

Check the report carefully, and please buy me a :coffee: , or better, a GPU :wink:!

github-actions[bot] avatar Dec 18 '23 11:12 github-actions[bot]