revise-code-for-catchability-with-bias-correction

nmfs-ost / ss3-source-code

The source code for Stock Synthesis (SS3).

https://nmfs-ost.github.io/ss3-website/

Creative Commons Zero v1.0 Universal

36 stars 16 forks source link

revise-code-for-catchability-with-bias-correction #544

Closed Rick-Methot-NOAA closed 8 months ago

Rick-Methot-NOAA commented 8 months ago

Resolves issue #535

What tests have been done?

<-- - [] Test files are in the issue. -->

What tests/review still need to be done?

needs close look by a math-stat whiz

Is there an input change for users to Stock Synthesis?

revision for T-dist

Rick-Methot-NOAA commented 8 months ago

Thanks Cole for the good suggestions. I originally was going to retain T-dist exactly as you suggest. The reason I went with the distribution you reviewed is that it allowed for options >=0 to use log(surveyobs) and all options <0 would be in normal space. I have never seen anyone use the T-dist (and wish we had a database of all applications so I would know!) , so I felt that the risk was low. But the risk is not zero, so I will consider you recommendation.

iantaylor-NOAA commented 8 months ago

I found an email from 2020 in which Max Cardinale (@akatan999) referenced using the t-distribution for index likelihoods. It came up because the index plots in r4ss were not working correctly for t-distributed indices. I think the bug was present for at least 6 months and since nobody else reported the issue during that time, it's possible that nobody else was using the t-distribution feature.

Rick-Methot-NOAA commented 8 months ago

Thanks Ian. I will maintain backward compatibility.

kellijohnson-NOAA commented 8 months ago

We are attempting to use the t-distribution for the age-1 index of Pacific Hake abundance in the 2024 stock assessment per your recommendations @Rick-Methot-NOAA when I talked to you last year about trying to fit estimates of zero.

kellijohnson-NOAA commented 8 months ago

Also, regarding my review of this code. I am right in the middle of an assessment so if you need my review it will have to wait until at least next week. Sorry.