A Critical Look at the Consistency of Causal Estimation with Deep Latent Variable Models
Details
Title : A Critical Look at the Consistency of Causal Estimation with Deep Latent Variable Models Author(s): Severi Rissanen, Pekka Marttinen
Summary
Paper runs experiments to study the consistency (of estimating causal effects) when deep latent variable models are used. For this, the authors use the CEVAE, and show that it does not estimate the causal effect correctly when latent variables are misspecified or the data distribution is complex - e.g. linear Gaussian data with an extra (proxy) variable with irrelevant high variance noise and linear Gaussian data with repeated proxies.