Abstract:Most dimension reduction models are suited for continuous but not for discrete covariates. A flexible model to incorporate both discrete and continuous covariates is to assume that some covariates, a q-dimensional vector Z, are related to the response variable, Y , through a linear relationship; while the remaining covariates, a p-dimensional vector X, are related to Y through k indices which are X ′B and some unknown function g. This results in a semiparametric model called semi-linear index model, which includes both the popular single-index model and the partial linear model as special cases. To avoid the curse of dimensionality, k should be much smaller than p, and this is often realistic as the key features of a high dimensional variable can often be extracted through a low-dimensional subspace. Two approaches to estimate the model components have been considered by Carroll, Fan, Gijbels and Wand (1997) and Yu and Ruppert (2002). Both focus on the simple case where k = 1, as these approaches are challenged computationally when k > 1. These computational challenges are partly triggered by the fact that the estimation of the parametric components is integrated with the nonparametric estimation of the link function g. Moreover, the theory for both approaches is incomplete. We show in this paper that a simple approach which separates the dimension reduction stage to estimate B from the remaining model components is available when the two covariates Z and X are independent. For instance, one can apply any suitable dimension reduction approach, such as the average derivative method, the projection pursuit regression or sliced inverse regression, to get an initial estimator for B which is consistent at the √ n rate, and then apply a suitable approach, such as the profile ∗Wang’s research was supported in part by NSF grants DMS-0406430 and DMS-98-03627. †Zhu’s research was supported by the grants (HKU7181/02H and HKU7060/04P) from The Research Grants Council of Hong Kong. AMS 2000 subject classifications: Primary 62G08, 62G20; secondary 62F12

A high-dimensional single-index regression for interactions between treatment and covariates

A Single-Index Model With a Surface-Link for Optimizing Individualized Dose Rules

A simple method for estimating interactions between a treatment and a large number of covariates

On efficient dimension reduction with respect to the interaction between two response variables

A dimension reduction based approach for estimation and variable selection in partially linear single-index models with high-dimensional covariates

Penalized Independence Rule for Testing High-Dimensional Hypotheses

Penalized Interaction Estimation for Ultrahigh Dimensional Quadratic Regression

Identifying sparse treatment effects in high-dimensional outcome spaces

Nonlinear Interaction Detection Through Model-Based Sufficient Dimension Reduction

Dimension reduction for integrative survival analysis

Strong Sure Screening of Ultra-high Dimensional Data with Interaction Effects

Impact of Sufficient Dimension Reduction in Nonparametric Estimation of Causal Effect

Semi-linear linear index model when the linear covariates and indices are independent

A New Covariate Selection Strategy for High Dimensional Data in Causal Effect Estimation with Multivariate Treatments

High-dimensional Interactions Detection with Sparse Principal Hessian Matrix

An Efficient Model‐free Approach to Interaction Screening for High Dimensional Data

On a Dimension Reduction Regression with Covariate Adjustment

Two-Sample Tests for High-Dimensional Linear Regression with an Application to Detecting Interactions.

A robust and efficient approach to causal inference based on sparse sufficient dimension reduction

On Dimension Reduction in Regressions with Multivariate Responses

Deep Neural Networks for Nonparametric Interaction Models with Diverging Dimension