Analysis Regression

Least-squares fit of metric_y on metric_x — the scatter + line view.

Fits an ordinary least-squares (OLS) straight line to the two metrics' aligned daily series and returns everything the Analysis page's Regression section draws and states:

• points: the aligned (x, y) scatter cloud.
• fit.line: the two-point OLS line; fit.lowess: a locally-weighted smoother showing the data's actual shape (so curvature is visible next to the straight line); fit.fitted/fit.residuals: the residual-vs-fitted diagnostic.
• fit: slope, intercept, Pearson r, R², the slope's 95% CI, a robust Theil–Sen slope cross-check, the Spearman r + is_linear linearity gate, and both the naive and autocorrelation-adjusted p-values.

lag shifts the response relative to the predictor (positive = y follows x) on the resample grid; resample="None" is rejected because a row shift over irregular raw timestamps is not a time shift. Regression is asymmetric — x is the predictor, y the response — so the pair order matters.