scientific-eda

Defensive exploratory data analysis captures problem context first, narrows to a single initial plot or summary, and prompts for clarification before executing any scripts on your scientific data files. This guided approach prevents premature code generation and ensures step-by-step validation.

Yes, you can analyze CSV, FASTA, and other scientific formats within a dedicated session structure. The process generates a per-session journal, saves WebP plots in a plots directory, and stores runnable Python scripts under an analysis folder.

Yes, uv is required to run Python scripts for this exploratory analysis workflow. Scripts are executed via uv-run using PEP 723 inline metadata, ensuring the environment and dependencies are managed directly within each script.

The best way to maintain an analysis journal is using a session-based structure that automatically records steps under an analysis directory. It captures problem context, tracks each plot or summary, and saves WebP visualizations alongside runnable scripts.

Defensive EDA differs by requiring human guidance before executing plots or summaries, asking 'why?' before generating output, and avoiding automatic file dumping. Standard approaches often open files and generate code immediately without capturing initial problem context.