Skip to content

Sufficient Dimension Reduction with Multimodal Predictor Objects

Abdul-Nasah Soale

stat.MEarXiv:2611.13460

Abstract

We study the regression of a scalar response Y on a Euclidean predictor X ∈ Rp and a random predictor object Z in a general metric space under arbitrary structural dependence. We introduce a joint sufficient dimension reduction (BX, GZ) constructed via a sequential conditional independence framework, which differs fundamentally from the ordinary central subspace of Y X. Methodologically, Z is embedded into a reproducing kernel Hilbert space, and a regularized kernel-response envelope model extracts a finite-dimensional reduction of the feature map. Partial adaptive estimation then recovers the Euclidean reduction conditional on the first-stage coordinates. We establish root-n consistency rates for the first-stage envelope subspace under a shrinking regularizer, alongside large-sample consistency for the second-stage Euclidean subspace estimator. Simulations and two real-data applications illustrate the utility and flexibility of the proposed method.

Create a lesson