correlation

correlation is a compact association check between two numeric columns. It is evidence for relationship, not a model of cause.

Use Pearson when a linear relationship is the question. Use Spearman when rank order or monotonic movement is more believable than raw linearity.

Association contract

Field	Required	Notes
`inputs`	yes	Exactly one table.
`columnA`, `columnB`	yes	Numeric columns to pair row-wise.
`method`	no	`pearson` by default; `spearman` ranks values first.

default reports method, paired n, coefficient, p-value, effect size, and a 95% coefficient confidence interval.

Pearson/Spearman disagreement can be more useful than either number alone because it often points to outliers or non-linear shape.

CORRELATION_SAMPLE_SMALL appears when n is below 20.
CORRELATION_PEARSON_OUTLIER_SENSITIVE appears when Pearson and Spearman differ by at least 0.2.
If you need a directional fitted relationship, move to linear_regression; if you need controls or nonlinear features, use Python/R.

- id: cor
  kind: correlation
  inputs: [data]
  columnA: x
  columnB: y
  method: pearson                 # pearson | spearman, default pearson