waveform-correlationlisted

# Waveform correlation (QC) Pairwise Pearson correlation across traces in a CloudSprite project: score each trace, flag outliers, describe a clean average of the rest. MCP in this plugin is **read-only**. You can find traces and reason about shape from summaries. You cannot publish an average dataset or an outlier notebook through MCP. ## Pearson r for waveform QC Pearson r is linear similarity of two y-arrays: −1 (anti-correlated) to +1 (same shape). | r | Meaning | |-|-| | ≥ 0.99 | Nearly identical | | 0.95–0.99 | Similar with natural variation | | 0.90–0.95 | Noticeable differences | | < 0.90 | Meaningfully different — likely an outlier | **Pairwise:** every trace vs every other. A trace's score = mean of its row in the n×n r matrix, **excluding the diagonal**. Outliers are more than σ standard deviations **below** the group mean (default σ = 2). Do **not** average first and then compare to the average — outliers pollute the reference. Score pairwise, then average only the good traces. ## Ask (or derive) before running ### 1. Project and trace type `search_datasets` / `get_dataset` in the bound project. List trace names actually present, then ask which to correlate. If only one type exists across the set, use it. ### 2. Which datasets A parameter (batch id), the whole project, or a name pattern. Confirm the filter and the count. ### 3. X-axis alignment If summaries show different x-ranges or point counts, ask: - **Interpolate** to a common grid (full span) —