Controlling for discrete unmeasured confounding in nonlinear causal models
This addresses a major challenge in causal inference for researchers and practitioners dealing with observational data, though it appears incremental as it builds on recent identifiability results.
The paper tackles the problem of identifying causal relationships from non-experimental data when discrete unmeasured confounding is present, by proposing a method that detects and corrects for such confounding under specific assumptions, with experimental results supporting its effectiveness.
Unmeasured confounding is a major challenge for identifying causal relationships from non-experimental data. Here, we propose a method that can accommodate unmeasured discrete confounding. Extending recent identifiability results in deep latent variable models, we show theoretically that confounding can be detected and corrected under the assumption that the observed data is a piecewise affine transformation of a latent Gaussian mixture model and that the identity of the mixture components is confounded. We provide a flow-based algorithm to estimate this model and perform deconfounding. Experimental results on synthetic and real-world data provide support for the effectiveness of our approach.