Module linear_regression 

Rolling least-squares linear regression

Fit (y = \text{intercept} + \text{slope}\cdot x) over the last N samples, where (x = 0,1,\ldots,N-1) (oldest → newest in the window).

Versatile: pass any f64 series — closes, highs, lows, typical price, volume, custom transforms. The crate does not force OHLCV; your engine picks the slice (e.g. highs for resistance slope, closes for trend slope).

---

Trading perspective

Output	Habit
slope	Direction & steepness per bar (price units / bar)
angle_degrees	atan(slope) in degrees — comparable trend “angle” when scale is fixed
r_squared	How linear the window is (1 = perfect line)
intercept	Fitted value at the oldest bar of the window

---

Engineering

Layer	API
Params	LinRegParams
Batch	linear_regression
Live	LinRegState::push / push_bars / from_history
Teaching	linear_regression_solution

Each push when warm is O(period) (recompute OLS on the ring). Fine for typical windows (20–200); not nanosecond-critical path.

Word problem

Closes 1,2,3,4,5 over five bars. Slope of the 5-bar regression?

Expect: slope 1.0 (perfect line).

use finance_solution::stocks::ta::{linear_regression, LinRegParams};
let y = [1.0, 2.0, 3.0, 4.0, 5.0];
let s = linear_regression(&y, LinRegParams::new(5)).unwrap();
assert!((s[4].unwrap().slope - 1.0).abs() < 1e-12);
assert!((s[4].unwrap().r_squared - 1.0).abs() < 1e-12);

Structs

LinRegBar

One fitted window.

LinRegParams

Rolling regression window length.

LinRegSolution

LinRegState

Incremental rolling regression on a caller-chosen series.

ValidatedLinReg

Validated pack.

Functions

linear_regression

Batch rolling OLS. series is your choice of bar field (close, high, …).

linear_regression_solution