Estimation and visualization of treatment effects for multiple outcomes

TL;DR

This paper introduces a multivariate regression approach with latent variables and Lasso constraints to identify and visualize subgroups with enhanced treatment effects across multiple outcomes, applicable to various data types.

Contribution

It proposes a novel multivariate regression method with latent variables and sparsity constraints for interpreting treatment effects on multiple outcomes simultaneously.

Findings

Effective identification of subgroups with better treatment responses.

Method applicable to various outcome types.

Demonstrated success on simulation and real data.

Abstract

We consider a randomized controlled trial between two groups. The objective is to identify a population with characteristics such that the test therapy is more effective than the control therapy. Such a population is called a subgroup. This identification can be made by estimating the treatment effect and identifying interactions between treatments and covariates. To date, many methods have been proposed to identify subgroups for a single outcome. There are also multiple outcomes, but they are difficult to interpret and cannot be applied to outcomes other than continuous values. In this paper, we propose a multivariate regression method that introduces latent variables to estimate the treatment effect on multiple outcomes simultaneously. The proposed method introduces latent variables and adds Lasso sparsity constraints to the estimated loadings to facilitate the interpretation of the relationship between outcomes and covariates. The framework of the generalized linear model makes it applicable to various types of outcomes. Interpretation of subgroups is made by visualizing treatment effects and latent variables. This allows us to identify subgroups with characteristics that make the test therapy more effective for multiple outcomes. Simulation and real data examples demonstrate the effectiveness of the proposed method.