A Critical Look at the Consistency of Causal Estimation With Deep Latent Variable Models
This work highlights a critical limitation in applying deep latent variable models for causal inference, which is important for researchers and practitioners relying on these methods for accurate causal estimates.
The paper investigates the consistency of causal estimation using deep latent variable models, focusing on CEVAE, and finds that it fails to correctly estimate causal effects under misspecified latent variables or complex data distributions, despite working in simple scenarios.
Using deep latent variable models in causal inference has attracted considerable interest recently, but an essential open question is their ability to yield consistent causal estimates. While they have demonstrated promising results and theory exists on some simple model formulations, we also know that causal effects are not even identifiable in general with latent variables. We investigate this gap between theory and empirical results with analytical considerations and extensive experiments under multiple synthetic and real-world data sets, using the causal effect variational autoencoder (CEVAE) as a case study. While CEVAE seems to work reliably under some simple scenarios, it does not estimate the causal effect correctly with a misspecified latent variable or a complex data distribution, as opposed to its original motivation. Hence, our results show that more attention should be paid to ensuring the correctness of causal estimates with deep latent variable models.