Analysis Planning

Use this skill when you need to design a scientific analysis pipeline
before writing or running any code. The planning phase ensures a sound
methodology before committing to implementation.

Planning Methodology

Follow these steps in order:

1. Understand the Question

• Restate the user's research question in your own words.
• Confirm any ambiguities before proceeding.

2. Survey the Data

• Examine available files, column names, units, and sample sizes.
• Note missing data, unexpected formats, or potential quality issues.

3. Design the Pipeline

Lay out each analysis step in order:

1. Data loading & parsing
2. Quality control checks (missing values, outliers, distributions)
3. Data transformations (normalization, filtering, alignment)
4. Primary analysis (statistical tests, model fitting, feature extraction)
5. Validation & sanity checks
6. Visualization & reporting

4. Specify Parameters

For each step, recommend:

• Which library / function to use
• Default parameter values with justification
• Expected output format and value ranges

5. Anticipate Risks

Flag potential pitfalls:

• What could go wrong at each step?
• What would invalidate the analysis?
• What fallback approaches exist?

6. Define Success Criteria

• What does a "good" result look like?
• How will you know the analysis worked correctly?

Incremental Execution Principle

Always plan for incremental validation:

1. Examine structure — load one representative file / sample first
2. Validate on one — run the full pipeline on a single sample
3. Small batch test — process 2–3 additional units, check consistency
4. Scale — only after steps 1–3 pass, process the full dataset

Output Format

Present the plan as a numbered checklist with clear deliverables at each
step. Include:

• Step name — concise label
• Action — what to do
• Tool / library — which package to use
• Expected output — what the result should look like
• Checkpoint — how to verify the step succeeded

Domain Customization