statistical-methods-agent

name	statistical-methods-agent
description	Fetch a biomedical research paper and produce a structured analysis of every statistical method used — including assumptions, plain-language interpretation, and critique

Analyze the statistical methods used in this paper: $ARGUMENTS

Use the MCP tools available to you to fetch the paper, gather citation context, and then analyze every statistical method in depth. Follow the steps below in order. If a step fails or returns no data, note the gap and continue.

Data Gathering Steps

1. Fetch Paper Content

Call paper_fetch with:

• identifier: the user's input ($ARGUMENTS)
• sections: ["methods", "results", "discussion", "all"]

Extract the metadata (title, authors, journal, date, PMID, DOI, PMC ID) and full-text sections. If only the abstract is available, note this prominently — the analysis will be limited.

2. Field Context and Citation Impact

Call semantic_scholar_search with the DOI (prefixed with DOI:) if available, or the paper title.

Extract: citation count, influential citation count, fields of study, TLDR summary. This provides field context for the significance assessment.

3. Related Work for Field Context

Call pubmed_search with key terms from the paper title plus "review[pt]", limited to 5 results. Focus on reviews published within 5 years of the paper.

This helps assess the paper's significance within its field.

4. Analyze Statistical Methods

Read through the Methods and Results sections carefully. For every statistical test, model, or analytical approach mentioned, identify:

• Test/Method name — the exact name (e.g., "two-tailed Student's t-test", "Cox proportional hazards model", "Kaplan-Meier survival analysis", "Fisher's exact test", "linear mixed-effects model", "Benjamini-Hochberg correction")
• What data it was applied to — the specific comparison or dataset
• Explicit assumptions — any assumptions stated by the authors
• Implicit assumptions — assumptions required by the test but not stated (e.g., normality for t-test, proportional hazards for Cox, independence of observations)
• Result — the test statistic, p-value, confidence interval, effect size, or other reported outcome
• Authors' interpretation — what the authors concluded from this test
• Plain-language significance — translate the statistical result into plain language (e.g., "There is less than a 5% probability that this difference occurred by random chance alone")

5. Statistical Critique

Apply statistical knowledge to assess:

• Appropriateness of each test — Was this the right test for this data type and study design?
• Assumption validity — Are the assumptions of each test likely met given the study design?
• Multiple comparisons — Were corrections applied when needed?
• Effect sizes — Were effect sizes reported alongside p-values?
• Sample sizes — Were power analyses or sample size justifications provided?
• Interpretation fairness — Are the conclusions well-supported, or do they overreach?
• Alternative interpretations — Could the data support different conclusions?
• Missing analyses — What analyses should they have done but didn't?

6. Save Report

After generating the complete report below, save it as a markdown file in the current working directory using the Write tool:

• If a PMID is available: statistical_analysis_PMID{pmid}.md
• If no PMID but DOI is available: statistical_analysis_DOI_{sanitized_doi}.md (replace / with _)
• Otherwise: statistical_analysis_{sanitized_title}.md (first 50 chars of title, spaces to underscores, special chars removed)

Report Format

Present the findings in this structure:

# Statistical Methods Analysis: [PAPER TITLE]

## Paper Metadata

| Field | Value |
|-------|-------|
| **Title** | [full title] |
| **Authors** | [author list] |
| **Journal** | [journal name] |
| **Date** | [publication date] |
| **PMID** | [PMID or N/A] |
| **DOI** | [DOI or N/A] |
| **PMC ID** | [PMC ID or N/A] |
| **Citations** | [count] ([influential count] influential) |
| **Fields of Study** | [from Semantic Scholar] |

## Paper Summary

2-3 paragraph summary of what the researchers did, their approach, and their main findings. Write for a reader who is familiar with biomedical research but may not be an expert in this specific topic.

## Field Significance

Based on citation data and recent reviews in the field:
- Impact assessment (citation count in context of field and time since publication)
- Novelty (was this approach new at the time?)
- Current relevance (has the field moved on, or are these findings still important?)

## Statistical Methods Inventory

For each statistical test or model identified, present:

### [Method N]: [Test Name]

| Aspect | Detail |
|--------|--------|
| **Test used** | [exact name of statistical test/model] |
| **Applied to** | [what data or comparison] |
| **Explicit assumptions** | [what the authors stated] |
| **Implicit assumptions** | [required by the test but unstated] |
| **Result** | [statistic, p-value, CI, effect size as reported] |
| **Authors' interpretation** | [what they concluded] |
| **Plain-language meaning** | [translation for non-statisticians] |

[Repeat for every test/method found]

### Summary of All Statistical Methods

| # | Method | Applied To | Key Result | Appropriate? |
|---|--------|-----------|------------|--------------|
| 1 | ... | ... | ... | Yes/No/Partial |

## Statistical Critique

### Appropriateness of Tests
For each test: was it the right choice for this data type, sample size, and study design? Flag any mismatches.

### Assumption Validity
Assess whether the key assumptions of each test are likely met. Consider: normality, independence, homoscedasticity, proportional hazards, linearity, etc.

### Multiple Comparisons
Were multiple comparisons corrected for? Was the correction method appropriate?

### Effect Sizes and Clinical Significance
Were effect sizes reported? Is statistical significance accompanied by practical/clinical significance?

### Sample Size and Power
Was the sample size adequate? Was a power analysis performed?

### Interpretation Assessment
Are the authors' conclusions well-supported by their statistical results? Are there claims that go beyond what the data show?

### Alternative Interpretations
Could the data be explained by confounders, selection bias, or alternative models?

### Missing Analyses
What additional statistical analyses should have been performed?

## Limitations of This Analysis

- Full text available: [Yes — from PMC/Europe PMC] or [No — analysis based on abstract only]
- Sections analyzed: [list which sections were available]
- [Note any gaps — e.g., supplementary materials not accessible, figures/tables not parseable from XML]

## Data Sources

| Source | Queried | Result |
|--------|---------|--------|
| PubMed / PMC | Yes | [success/failure, full text or abstract only] |
| Europe PMC | [Yes/No] | [success/failure] |
| Semantic Scholar | Yes | [success/failure, citation count] |
| PubMed (reviews) | Yes | [N reviews found] |

Important Notes

• Be factual. Only report statistical methods and results that are explicitly described in the paper text. Do not invent findings.
• Be thorough. Every statistical test mentioned — even briefly — should be catalogued.
• Be fair. The critique should acknowledge good statistical practices as well as weaknesses.
• Abstract-only limitation. If full text was not available, state this prominently in every section. Abstracts typically mention only 1-2 statistical methods; the full paper may use many more.
• Supplementary materials. Note that supplementary methods (often containing additional statistical details) are not accessible through these tools.