跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
其他技能

data-explorer

Descriptive analysis and EDA on cleaned data: distributions, missingness mechanism screen, correlations, categorical structure. No p-values for group comparisons.

MAGA2010 · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add MAGA2010/data-master --skill data-explorer

技能说明


name: data-explorer description: "Descriptive analysis and EDA on cleaned data: distributions, missingness mechanism screen, correlations, categorical structure. No p-values for group comparisons." parent: data-master stage: 2a

data-explorer

Stage 2a of the data-master pipeline. The router in ../../SKILL.md owns the end-to-end run; this skill owns one stage and one gate. Read the parent contract before changing anything here: the invariants in ../../SKILL.md (additive-only, types decided once, default-FAIL gates, three-part provenance) are not restated, they are inherited.

When to invoke

After cleaning passes its gate. Run this before inferential statistics so the test choice is informed by the actual distributions, not assumed.

Inputs

  • outputs/cleaned/cleaned.csv via dm_types.load_cleaned (manifest replayed)
  • analysis/profile.json

Outputs

  • analysis/descriptive.json: numeric (mean, SD, SE, median, IQR, range, skew, missing_pct), categorical (levels, counts, mode), missing_mechanism_screen, correlation_pearson

Commands

python scripts/stats_descriptive.py --clean outputs/cleaned/cleaned.csv --out outputs/analysis/descriptive.json

Gate (default FAIL)

Every numeric variable carries n, missing_pct, mean and median, SD, IQR and skew. Reporting only the mean for a strongly skewed variable is a fail, because the mean is then the least trustworthy number on the page.

Failure modes seen in real runs

  • Mean without median hides a 3 SD outlier.
  • A missingness screen presented as a mechanism proof. Little's test describes the pattern; it cannot prove MCAR. The wording must say "screening result".
  • Pearson only on a monotone-but-curved pair understates the association; Spearman is reported alongside, and its use is recorded.

Do not

  • do not compute group p-values here -- that is the statistical-analyzer's job;
  • do not read the cleaned file with pd.read_csv; that re-guesses the types;
  • do not drop the median because the mean was requested.

Upstream / downstream

  • previous: ../data-cleaner/SKILL.md
  • next: ../statistical-analyzer/SKILL.md
  • contract: ../../references/workflow-contract.md