Module concurrent_regression 

Concurrent (varying-coefficient) functional regression.

Models the relationship:

y_i(t) = β₀(t) + Σₖ βₖ(t)·xₖᵢ(t) + εᵢ(t)

where all functional curves share the same dense grid of m evaluation points. Estimation follows the fdaconcur ptFCReg → smPtFCRegCoef two-step convention:

1. Pointwise OLS at each grid column j — build a design matrix [1 | x₁[:,j] | … | xₚ[:,j]] and solve the normal equations to obtain raw coefficient estimates raw_β₀[j] and raw_βₖ[j] for each predictor.
2. Local-linear kernel smoothing of each raw discrete coefficient sequence over j via crate::smoothing::local_linear. The bandwidth parameter is the sole roughness knob: larger bandwidth → smoother β(t).

Boundary behaviour: the smoother is not edge-corrected; the first and last few grid points exhibit higher bias. Tests check only interior indices.

Structs

ConcurrentRegrResult

Result of concurrent (varying-coefficient) functional regression.

Functions

concurrent_regression

Concurrent (varying-coefficient) functional regression.