Weighted quantile regression for one mixture component

Solves argmin_b sum_i w_i rho_tau(y_i - x_i' b) via quantreg::rq.wfit().

Usage

weighted_rq(X, y, tau, w, beta_prev = NULL)

Arguments

X

Design matrix (with intercept column).

y

Response vector.

tau

Quantile level.

w

Non-negative observation weights (the responsibilities p[, j]).

beta_prev

Optional fallback coefficients if the solve fails.

Value

A coefficient vector of length ncol(X).

Details

Exported as part of the mixqr extension API: building blocks that companion packages (e.g. location-varying gating, non-crossing) reuse to build custom EM engines on the same component machinery.