lingzhi227/agent-research-skills

data-analysis

Generate statistical analysis code with 4-round review.

ソースを見る

使い方

  1. インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
  2. 必要なツール、認証情報、対応環境を原文で確認します。
  3. 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/lingzhi227/agent-research-skills --skill data-analysis

開始前の確認

  • 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
  • スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文

見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。

Data Analysis

Generate rigorous statistical analysis code with multi-round review.

Input

  • $0 — Data source (CSV, JSON, pickle, or experiment logs)
  • $1 — Research goal or hypothesis to test

References

  • 4-round code review prompts: ~/.claude/skills/data-analysis/references/review-prompts.md

Scripts

Statistical summary and comparison

bash
python ~/.claude/skills/data-analysis/scripts/stat_summary.py --input results.csv --compare method --metric accuracy --output summary.json
python ~/.claude/skills/data-analysis/scripts/stat_summary.py --input results.csv --describe

Detects data types, recommends tests, runs comparisons, outputs effect sizes and significance stars. Requires numpy, scipy.

Format p-values

bash
python ~/.claude/skills/data-analysis/scripts/format_pvalue.py --values "0.001 0.05 0.23" --format stars
python ~/.claude/skills/data-analysis/scripts/format_pvalue.py --csv results.csv --column pvalue --format latex

Formats p-values with stars, LaTeX notation, or plain text. Stdlib-only.

Workflow

Step 1: Generate Analysis Code

Structure the code with these sections:

  1. # IMPORT — pandas, numpy, scipy, statsmodels, sklearn
  2. # LOAD DATA — Load from original data files
  3. # DATASET PREPARATIONS — Missing values, units, exclusion criteria
  4. # DESCRIPTIVE STATISTICS — Summary tables if needed
  5. # PREPROCESSING — Dummy variables, normalization
  6. # ANALYSIS — Statistical tests per hypothesis
  7. # SAVE ADDITIONAL RESULTS — Extra results to pickle

Step 2: 4-Round Code Review

  1. Round 1 — Code Flaws: Mathematical/statistical errors, wrong calculations, trivial tests
  2. Round 2 — Data Handling: Missing values, units, preprocessing, test choice
  3. Round 3 — Per-Table: Sensible values, measures of uncertainty, missing data
  4. Round 4 — Cross-Table: Completeness, consistency, missing variables

Step 3: Produce Results

  • Every nominal value must have uncertainty (CI, STD, or p-value)
  • Statistical tests must be appropriate for the data type
  • Results must match actual data — never hallucinate

Allowed Packages

pandas, numpy, scipy, statsmodels, sklearn, pickle

Statistical Test Selection

Data TypeTest
Two groups, normalIndependent t-test
Two groups, non-normalMann-Whitney U
Paired samplesPaired t-test / Wilcoxon
Multiple groupsANOVA / Kruskal-Wallis
CategoricalChi-square / Fisher's exact
CorrelationPearson / Spearman
RegressionOLS / Logistic / Mixed effects

Rules

  • Always report p-values for statistical tests
  • Account for relevant confounding variables
  • Use inherent package functionality (e.g., formula = "y ~ a * b" for interactions)
  • Do not manually implement available statistical functions
  • Access dataframes using string-based column names, not integer indices

Related Skills

タスクに関連する項目

同じリポジトリから

関連する Skills

すべての Skills
lingzhi227
コミュニティ

backward-traceability

Make every number in the final PDF traceable to the exact code line that produced it. Uses \hypertarget/\hyperlink LaTeX commands and \num{formula} evaluated at compile time. Use for reproducibility and data integrity verification.

1時間の導入数
2
GitHub Stars
451
更新日
2月27日
lingzhi227
コミュニティ

experiment-code

Write ML experiment code with iterative improvement. Generate training/evaluation pipelines, debug errors, and optimize results through code reflection. Use when implementing experiments for a research paper.

1時間の導入数
1
GitHub Stars
420
更新日
2月27日
lingzhi227
コミュニティ

paper-assembly

Orchestrate the full paper pipeline end-to-end. Manage state propagation between phases (literature → plan → code → experiments → figures → tables → writing → review), support checkpointing and resumption. Use for assembling a complete paper from components.

1時間の導入数
1
GitHub Stars
420
更新日
2月27日
lingzhi227
コミュニティ

literature-review

Conduct comprehensive literature reviews using multi-perspective dialogue simulation. Generate diverse expert personas, conduct grounded Q&A conversations, and synthesize findings into structured knowledge. Use when starting a new research project or writing a survey section.

1時間の導入数
1
GitHub Stars
414
更新日
2月27日