[feat]: get index error distribution from SS3 models

未关闭
  • 哪里不对？
#471 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
54/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
r
领域
data-visualization

调研方向

Read convert_output() and plot_index() to trace how $cpue becomes the plotted index intervals; the issue does not name source files or tests. Inspect how r4ss::SS_output() provides $survey_error and $survey_units, then decide how that information should be represented in the converted output. Done means index intervals reflect each SS3 error structure, including symmetric intervals for normal errors.

由索引模型根据 Issue 内容生成。

描述

enhancement
Is your feature request related to a problem? Please describe.

Thanks to @ericward-noaa for adding index intervals in https://github.com/nmfs-ost/stockplotr/pull/468/. This is an important feature that is important to be able to interpret the index fits.

One refinement needed in the future is to account for the fact that SS3 has multiple options for error structure for indices:

  • -1=normal;
  • 0=lognormal;
  • 1=lognormal with bias correction;
  • >1=df for T-distribution

Of these, lognormal probably covers at least 95% of models, but the normal error is used for models that include an environmental predictor of recruitment deviations.

The plot below was created loading stockplotr from Eric's fork associated with #468 and running the following code which relies on a petrale sole model with an index of recruitment deviations posted by @okenk. The "env" fleet uses normal error since the index spans zero so the intervals should be symmetric around the points and not tiny for observations close to zero.

petrale <- r4ss::SS_output('https://raw.githubusercontent.com/okenk/recruit_index_sims/refs/heads/main/simulations/petrale/0.2_30/30/em/')
petrale_converted <- convert_output(petrale)
plot_index(petrale_converted, interval = 0.95, module = "cpue")
Image
Describe the solution you would like.

I don't think that convert_output() is currently getting the information on error structure. In Report.sso that's in a separate table from the main table of index output.

Currently, r4ss::SS_output() returns the table $cpue (which is converted and serves the basis for the plots) as well as two vectors $survey_units (which is probably not needed for plotting), and $survey_error, which contains the codes above.

I'm happy to modify r4ss::SS_output() to merge those vectors into $cpue so that units and error were just additional columns. These would then need to be somehow integrated into the table created by convert_output(), ideally in a format that is similar to what FIMS will use to store index error structure assuming that in the future there will be additional uncertainty options beyond the current ~ dlnorm() option.

Describe alternatives you have considered

stockplotr::convert_output() could also do the merging of the separate SS3 tables

Additional context

No response

主要语言
R
星标
20
派生
10
平均合并
3 天 8 小时
30 天内合并 PR
31

环境准备

在 Codespaces 中打开

在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

nmfs-ost/stockplotr 的其他 Issue

查看 nmfs-ost/stockplotr 的全部 Issue

相似的 Issue

更多 R Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。