A Critical Look At The Identifiability of Causal Effects with Deep Latent Variable Models

Using deep latent variable models in causal inference has attracted considerable interest recently, but an essential open question is their identifiability. While they have yielded promising results and theory exists on the identifiability of some simple model formulations, we also know that causal effects cannot be identified in general with latent variables. We investigate this gap between theory and empirical results with theoretical 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 identify the correct causal effect with a misspecified latent variable or a complex data distribution, as opposed to the original goals of the model. Our results show that the question of identifiability cannot be disregarded, and we argue that more attention should be paid to it in future work.

[1]  J. Pearl,et al.  Measurement bias and effect restoration in causal inference , 2014 .

[2]  Graham Neubig,et al.  Lagging Inference Networks and Posterior Collapse in Variational Autoencoders , 2019, ICLR.

[3]  Nigam H. Shah,et al.  Counterfactual Reasoning for Fair Clinical Risk Prediction , 2019, MLHC.

[4]  David M. Blei,et al.  Adapting Neural Networks for the Estimation of Treatment Effects , 2019, NeurIPS.

[5]  Jennifer L. Hill,et al.  Bayesian Nonparametric Modeling for Causal Inference , 2011 .

[6]  Uri Shalit,et al.  Estimating individual treatment effect: generalization bounds and algorithms , 2016, ICML.

[7]  Jean-Philippe Vert,et al.  MissDeepCausal: Causal Inference from Incomplete Data Using Deep Latent Variable Models , 2020, ArXiv.

[8]  Daan Wierstra,et al.  Stochastic Backpropagation and Approximate Inference in Deep Generative Models , 2014, ICML.

[9]  Uri Shalit,et al.  Identifying Causal Effect Inference Failure with Uncertainty-Aware Models , 2020, NeurIPS.

[10]  Z. Geng,et al.  Identifying Causal Effects With Proxy Variables of an Unmeasured Confounder. , 2016, Biometrika.

[11]  Daniel C. Castro,et al.  Deep Structural Causal Models for Tractable Counterfactual Inference , 2020, NeurIPS.

[12]  Ruocheng Guo,et al.  Linked Causal Variational Autoencoder for Inferring Paired Spillover Effects , 2018, CIKM.

[13]  Kyungwoo Song,et al.  Counterfactual Fairness with Disentangled Causal Effect Variational Autoencoder , 2020, AAAI.

[14]  David M. Blei,et al.  The Blessings of Multiple Causes , 2018, Journal of the American Statistical Association.

[15]  Toniann Pitassi,et al.  Fairness through Causal Awareness: Learning Causal Latent-Variable Models for Biased Data , 2018, FAT.

[16]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[17]  Max Welling,et al.  Causal Effect Inference with Deep Latent-Variable Models , 2017, NIPS 2017.

[18]  J. Pearl,et al.  Causal Inference in Statistics: A Primer , 2016 .

[19]  Judea Pearl,et al.  The seven tools of causal inference, with reflections on machine learning , 2019, Commun. ACM.

[20]  Mihaela van der Schaar,et al.  GANITE: Estimation of Individualized Treatment Effects using Generative Adversarial Nets , 2018, ICLR.

[21]  Ali Razavi,et al.  Preventing Posterior Collapse with delta-VAEs , 2019, ICLR.

[22]  David Wipf,et al.  The Usual Suspects? Reassessing Blame for VAE Posterior Collapse , 2019, ICML.

[23]  Joshua D. Angrist,et al.  Identification of Causal Effects Using Instrumental Variables , 1993 .

[24]  Kun Zhang,et al.  Inferring Personalized and Race-Specific Causal Effects of Genomic Aberrations on Gleason Scores: A Deep Latent Variable Model , 2020, Frontiers in Oncology.