Quantile regression characterizes heterogeneous covariate effects across the response distribution, but its extension to high-dimensional longitudinal compositional data poses substantial challenges, such as the nonsmooth check loss, the unit-sum structure of compositional covariates. Existing work is confined to Lasso penalization, and does not readily extend to nonconvex penalties. We propose CDaLCqr, a unified coordinate descent algorithm for penalized longitudinal log-contrast quantile regression accommodating popular penalties, such as adaptive Lasso, smoothly clipped absolute deviation (SCAD), and minimax concave penalty (MCP) within a single framework. The proposed method uses a Huber-type smooth approximation of the check loss and enforces the zero-sum constraint through an augmented Lagrangian formulation. CDaLCqr combines coordinate descent with majorization-minimization, yielding explicit updates through coordinate-wise quadratic surrogates. We show that the stopping tolerance controls errors in the zero-sum constraint and stationarity conditions. Under suitable conditions and sufficiently close initialization, we also prove local convergence of the coefficient and multiplier iterates, with an error bound decreasing by a fixed factor less than one each iteration. Simulations highlight the estimation advantages of SCAD and MCP variants and their sparse model identification. An application to Chinese provincial industrial data identifies interpretable, quantile-specific associations between sectoral composition and regional economic output.

翻译：暂无翻译