Dealing with chains of different length

未关闭
  • 哪里不对？
#334 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

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
42/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
停滞
技术栈
r
领域
data

调研方向

首先比较 repair_draws() 创建不等链长之后的 niterations，以及 draws_df、draws_matrix、draws_list 和 draws_array 的转换入口。将 merge_chains 作为建议的解决方案进行检查，并确定应在哪里产生信息性警告或错误。当不等链得到一致处理，并且 draws_array 在需要时将用户指向 merge_chains，即表示完成。

由索引模型根据 Issue 内容生成。

描述

I suppose there is an implicit expectation that all chains are the same length, but different lengths can occur when doing filtering with e.g. dplyr. I think the recommended workflow should likely be to merge chains before doing this kind of filtering, but I guess that can't be enforced.

Currently, different draws formats handle chains of varying length differently in how they calculate niterations (even after repairing). And draws_array does not allow different length chains at all -> I think the error message could point to merge_chains as a solution.

Consider filtering a draws_df based on the value of some variable (below dplyr is used to do so, but it could be done in base R syntax). This can end up with a draws_df that has different length chains.

library(posterior)
library(dplyr)
fdraws <- example_draws() |>
as_draws_df() |>
dplyr::filter(`mu` > 0) |>
repair_draws()

niterations(fdraws)
## [1] 94

niterations(as_draws_matrix(fdraws))
## [1] 89.5

niterations(as_draws_list(fdraws))
## [1] 90

as_draws_array(fdraws)
## Error in abind::abind(x, along = 3L) : 
##   arg 'X2' has dims=86, 10, 1; but need dims=90, 10, X

Should there be a message/warning given when a draws object has different length chains that suggests merge_chains?

主要语言
R
星标
172
派生
27
平均合并
5 天 12 小时
30 天内合并 PR
5

环境准备

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。

stan-dev/posterior 的其他 Issue

查看 stan-dev/posterior 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。