Simultaneous variable selection and smoothing for high-dimensional function-on-scalar regression (Q1711594)

The paper under review deals with the problem of simultaneously selecting important predictors and producing smooth estimates of their effects in a function-on-scalar linear regression model (where the parameters and errors lie in a real separable Hilbert space but the predictors are real-valued scalars) with a large number of scalar predictors. According to the authors: ``While other methods are available for selection and estimation, none of them incorporate smoothing as well.'' The authors solve the cited problem by presenting a new functional linear adaptive mixed estimation (FLAME) methodology. They also provide a fast algorithm for computing the estimators, which is based on a functional coordinate descent, and an R package that is customized for functional data, resulting in substantial gains in computational efficiency. Asymptotic properties of the estimators are developed and simulations are provided to illustrate the advantages of FLAME over existing methods, both in terms of statistical performance and computational efficiency. The paper is concluded with an application to childhood asthma, where a potentially important genetic mutation is found that was not selected by previous functional data based methods, and with the discussion about opportunities for improvement.