Add constraints to deviation generators

[Bai-Li-NOAA]

ADMB stock assessment models add constraints to their recruitment deviation estimates. For example, the mean and sum of the recruitment deviations from BAM and ASAP are 0. However, the operating model from the model comparison project uses rnorm without constraints. The mean and sum of the randomly generated recruitment deviations will not be zero. The comparisons between OM and EMs could be found here: https://github.com/Bai-Li-NOAA/homework/blob/main/recruitment_deviations_check/recdev_check.md

To do:

• Add arguments to r_dev_case and f_dev_case functions to allow adding constraints to rnorm outputs
• Run OM with constraints and EMS. Get relative errors in MSY, FMSY, and SSBMSY. Expect lower relative errors compared to model comparison paper results

Thanks Kyle Shertzer for pointing out the issue!

[k-doering-NOAA]

@Bai-Li-NOAA , I'm not sure this is relevant , but I know there is some sort of bug with using sum-to-zero vectors in admb with MCMC. I'm pretty sure you are doing everything in R, but just in case thought I would mention!

[iantaylor-NOAA]

Here's the ADMB issue: admb-project/admb#107.
@Bai-Li-NOAA, it sounds like you're just talking about adding zero-sum vectors as an option in the OM, so can safely ignore this issue.

[Bai-Li-NOAA]

Thanks @k-doering-NOAA and @iantaylor-NOAA for mentioning the ADMB issue. I learn something new every time.

Yes, I am modifying the operating model in R, so no worries there.