TO DO: Confirm that new tagging data fixes proportions to have no negative numbers

[brianlangseth-NOAA]

Currently, the number of recaptures is greater than number of releases. So for tags not recaptured there can be negative values. YFT is currently being updated, so close this issue once confirm the recap data is right.

[brianlangseth-NOAA]

Data were updated and provided on Jan 27 but there remains issues with the naming convention for fleets. Consequently, the change in commit c18355a required more work. More recent commits (3addb3c) are placeholders to use old data, until the second fix is out

[brianlangseth-NOAA]

The above commit (02276b6) updates the EM.dat file to include the new tagging data, and accounts for the still present, but very small and few, negative numbers. These negative numbers are known and are an artefact of the SPM sampling routine. Changes to the data may still occur.

From Aaron Berger's email when sending the data on Jan 27

Hi everyone,

We are providing an update to the simulated datasets previously provided on github, to correct some problems recently identified by Matt Vincent, Rick Methot, and Haikun Xu. Recall that the github repo has switched locations, it is now here. The new datasets are located in ./data/Datasets_current_UseThese. In summary:

1. There were two errors in compiling the tag recovery data, which together led to some cases in which there were more recoveries than releases, and the apparent truncation of the tag recoveries at the age of 20 (Issue #6 on github). These errors have now been corrected. Please note that, due to the SPM algorithm, there are still a few cases (release groups) with low release numbers where slightly more tags are recovered than released. This is not a data generation error, but please be aware in case it causes a problem for your software, in which case it’s reasonable to adjust the data (e.g., drop the release group).
2. There was an error in compiling the various LF replicates, which resulted in the 221-cell LF data not matching the respective 1-area and 4-area compiled datasets. This error has now been fixed.
3. There was also a minor inconsistency between the first (19 July 2021) and second (15 Sept 2021) versions of the data in the way the LF data were generated, which slightly affected the fine scale 221-cell observations. The first version of the data has been replaced to remove this inconsistency.

The formats of the new datasets are identical to the earlier versions, so you should not have to change any analysis code. Please let us know (either by posting to github or by email) if you find any more problems, and we’ll address them ASAP.

[brianlangseth-NOAA]

Email from Aaron Berger on March 8 updated data (hopefully) a final time. Changes were small ones to tagging prop and to survey.

Dear participants in the Spatial Stock Assessment Simulation Project,

Please read through the entirety of this email as there are important updates, reminders, and a request for your feedback.

New Datasets

First off, thank you very much for your patience as we worked through some unforeseen issues. We have now corrected and updated the simulated datasets for Indian Ocean yellowfin tuna and uploaded them to github. You can find them in the main github repo under the data folder here.

Changes Made

In summary, we have fixed several problems, but the main issue was that many of the simulated datasets included release groups (year + age class) where more tags were recovered than had been released. Thanks to Matt Vincent and Rick Methot for identifying this problem. This problem occurred because SPM input files specify the number of fish scanned for tags in each stratum, and the simulation inputs did not provide the correct numbers. Providing these numbers required code changes to generate a new set of reports from the SPM operating model and multiple runs to evaluate and test the data generation procedures - all which took time.

The formats of the new datasets are largely the same as the earlier versions, so you likely will not have to make any significant change to your analysis code. Please let us know (either by posting to github or by email) if you have any questions/comments.

Reminders

1. The main dataset uses an effective sample size (at the 221-cell level) for length frequency data of 5 (labelled "....ESS_05.RData"). Two additional effective sample size datasets have also been provided for your exploratory pleasure (one with it set to 25: "....ESS_25.RData"; and one at the nominal level: "....ESS_00x.RData", where the nominal is the number of fish measured in the original dataset). There were 100 simulations at each ESS, which generated data at the 221-cell level.

2. The data are provided at three spatial scales: 1 area, 4 area, and 221 cell. The 1-area and 4-area datasets have been aggregated from the 221-cell datasets and are the main spatial areas to examine. The 221-cell dataset is provided so that you can explore the data at the finest scale and run additional analysis at that scale as desired.

3. The YFT_SRD_1A_4.RData and YFT_SRD_4A_4.RData files are provided as the one area and four area test datasets, respectively, and represent simulation dataset number 4 out of 100 with ESS=05. These datasets are provided so that everyone 'builds' their model(s) using a single representative dataset (SRD).