correlation¶
Compute a pairwise correlation matrix over the numeric columns of a file or
open tab. Read-only analytics (stays available under --mcp-read-only).
When to use¶
- Spot related numeric features before modelling.
- Quick "which columns move together?" check.
Input schema¶
| Parameter | Type | Required? | Default | Description |
|---|---|---|---|---|
path |
string | yes* | (no default) | Path to the file (omit when open_tab is set) |
open_tab |
string | no | (no default) | Operate on an open GUI tab (@active or a tab name) |
table |
string | no | (no default) | Specific table for multi-table sources |
method |
string | no | pearson |
pearson (linear) or spearman (monotonic, rank-based) |
unlimited |
bool | no | false |
Lift the 5,000,000-row file-loader cap so every row is used |
Non-numeric columns are ignored. For each pair, only rows where both values are present are used.
Response shape¶
{
"columns": ["height", "weight", "age"],
"matrix": [
[1.0, 0.82, 0.15],
[0.82, 1.0, 0.10],
[0.15, 0.10, 1.0]
]
}
matrix[i][j] is the correlation of columns[i] with columns[j]. A
coefficient is null when undefined (fewer than two paired rows, or zero
variance).
Example call¶
See also¶
profile: per-column summary statistics.value_frequency: distribution of a single column.