Abstract:In multivariate functional data analysis, different functional covariates can be homogeneous. The hidden homogeneity structure is informative about the connectivity or association of different covariates. The covariates with pronounced homogeneity can be analyzed jointly within the same group, which gives rise to a way of parsimoniously modeling multivariate functional data. In this paper, a novel grouped multivariate functional regression model with a new regularization approach termed "coefficient shape alignment" is developed to tackle the potential homogeneity of different functional covariates. The modeling procedure includes two main steps: first detect the unknown grouping structure with the new regularization approach to aggregate covariates into disjoint groups; and then the grouped multivariate functional regression model is established based on the detected grouping structure. In this new grouped model, the coefficient functions of covariates in the same homogeneous group share the same shape invariant to scaling. The new regularization approach builds on penalizing the discrepancy of coefficient shape. The consistency property of the detected grouping structure is thoroughly investigated, and the conditions that guarantee uncovering the underlying true grouping structure are developed. The asymptotic properties of the model estimates are also developed. Extensive simulation studies are conducted to investigate the finite-sample properties of the developed methods. The practical utility of the proposed methods is illustrated in the real data analysis on sugar quality evaluation. This work provides a novel means for analyzing the underlying homogeneity of functional covariates and developing parsimonious model structures for multivariate functional data.

Subgroup analysis for high-dimensional functional regression

Subgroup Analysis for the Functional Linear Model

Partially Functional Linear Regression in High Dimensions

High dimensional test for functional covariates

Regression‐based heterogeneity analysis to identify overlapping subgroup structure in high‐dimensional data

Functional Linear Regression with Mixed Predictors

Coefficient Shape Alignment in Multiple Functional Linear Regression

Subgroup identification in clinical trials: an overview of available methods and their implementations with R

Heterogeneous quantile regression for longitudinal data with subgroup structures

Regression and Subgroup Detection for Heterogeneous Samples

Coefficient Shape Alignment in Multivariate Functional Regression

Subgroup Analysis for Longitudinal data via Semiparametric Additive Mixed Effect Model

Regularized partially functional quantile regression

Variable Selection and Minimax Prediction in High-dimensional Functional Linear Model

ROBUST SUBGROUP IDENTIFICATION

Regression Analysis of Asynchronous Longitudinal Functional and Scalar Data

Clusterwise Functional Linear Regression Models.

Subgroup analysis for longitudinal data based on a partial linear varying coefficient model with a change plane

Regression with Functional Errors-in-Predictors: A Generalized Method-of-Moments Approach

Data-Driven Subgroup Identification for Linear Regression

Functional variable selection via Gram–Schmidt orthogonalization for multiple functional linear regression