Abstract:Estimating treatment effects over time is relevant in many real-world applications, such as precision medicine, epidemiology, economy, and marketing. Many state-of-the-art methods either assume the observations of all confounders or seek to infer the unobserved ones. We take a different perspective by assuming unobserved risk factors, i.e., adjustment variables that affect only the sequence of outcomes. Under unconfoundedness, we target the Individual Treatment Effect (ITE) estimation with unobserved heterogeneity in the treatment response due to missing risk factors. We address the challenges posed by time-varying effects and unobserved adjustment variables. Led by theoretical results over the validity of the learned adjustment variables and generalization bounds over the treatment effect, we devise Causal DVAE (CDVAE). This model combines a Dynamic Variational Autoencoder (DVAE) framework with a weighting strategy using propensity scores to estimate counterfactual responses. The CDVAE model allows for accurate estimation of ITE and captures the underlying heterogeneity in longitudinal data. Evaluations of our model show superior performance over state-of-the-art models.

What problem does this paper attempt to address?

This paper attempts to solve two main problems encountered when estimating individual treatment effects (ITE) in longitudinal data: unobserved adjustment variables and time - varying effects. Specifically: 1. **Unobserved adjustment variables**: In many real - world applications, such as precision medicine, epidemiology, economics, and marketing, it is crucial to estimate the change in treatment effects over time. However, due to the existence of unobserved confounding factors or risk factors that only affect the outcome sequence, it becomes difficult to accurately estimate the causal effect. These problems are particularly prominent in longitudinal data because the data usually involves repeated measurements and the treatment effects may vary at different time points. 2. **Time - varying effects**: In longitudinal data, treatment effects may change over time. To accurately estimate ITE, it is crucial to capture this dynamic nature. In addition, unobserved adjustment variables introduce additional heterogeneity, making the estimation more complex. To solve these problems, the paper proposes the **Causal Dynamic Variational Auto - Encoder (Causal DVAE, CDVAE)** model. This model combines the Dynamic Variational Auto - Encoder (DVAE) framework and a propensity - score - based weighting strategy to estimate counterfactual responses. The main contributions of CDVAE include: - **Handling unobserved adjustment variables**: By learning latent adjustment factors, it captures hidden changes in treatment responses. - **Capturing time dynamics**: Using the Seq2Seq architecture to automatically encode the responses of interest, thereby modeling the complex relationships and dependencies in longitudinal data. - **Theoretical validity**: Assuming a conditional Markov model, it ensures the validity of the learned adjustment variables and proves that ITE can be identified from the data given these representations. In conclusion, this paper aims to provide a more accurate method for estimating individual treatment effects in longitudinal data through the CDVAE model, while taking into account unobserved heterogeneity and time - varying effects.

Causal Dynamic Variational Autoencoder for Counterfactual Regression in Longitudinal Data

DR-VIDAL -- Doubly Robust Variational Information-theoretic Deep Adversarial Learning for Counterfactual Prediction and Treatment Effect Estimation on Real World Data

Causal Effect Estimation using identifiable Variational AutoEncoder with Latent Confounders and Post-Treatment Variables

Variational Causal Inference

Leaning Time-Varying Instruments for Identifying Causal Effects in Time-Series Data

Causal Effect Estimation with Variational AutoEncoder and the Front Door Criterion

Causal Contrastive Learning for Counterfactual Regression Over Time

Causal Inference with Conditional Front-Door Adjustment and Identifiable Variational Autoencoder

Variational Temporal Deconfounder for Individualized Treatment Effect Estimation from Longitudinal Observational Data

Causal Inference using Multivariate Generalized Linear Mixed-Effects Models with Longitudinal Data

Counterfactual Generative Modeling with Variational Causal Inference

Disentangled Representation via Variational AutoEncoder for Continuous Treatment Effect Estimation

EDVAE: Disentangled Latent Factors Models in Counterfactual Reasoning for Individual Treatment Effects Estimation

Learning Conditional Instrumental Variable Representation for Causal Effect Estimation

CDVD: Causal Dynamic Variational Deconfounder for Estimating Parameter Adjusting Effect

Deep Bayesian Estimation for Dynamic Treatment Regimes with a Long Follow-up Time

A Variational Autoencoder for Heterogeneous Temporal and Longitudinal Data

Counterfactual Generative Models for Time-Varying Treatments

Estimating Individual Dose-Response Curves under Unobserved Confounders from Observational Data

VACA: Design of Variational Graph Autoencoders for Interventional and Counterfactual Queries

VACA: Designing Variational Graph Autoencoders for Causal Queries