Search papers, labs, and topics across Lattice.
This paper introduces a novel method for falsifying candidate causal graphs by assessing their ability to explain the propagation of outlier events, leveraging the principle that weak outliers rarely cause strong ones. By turning this principle on its head, the authors develop statistical tests that evaluate whether a proposed causal graph is consistent with observed outlier data, providing a framework for causal inference in the absence of ground truth. The results demonstrate that these tests maintain false positive control and power guarantees, even when only a single outlier sample is available.
Outlier events can be a powerful tool for falsifying causal graphs, revealing inconsistencies that traditional methods might miss.
True causal relationships are rarely known, and inferring causal graphs from data is hard. A fundamental challenge is how to assess whether a given causal graph is good in the absence of a ground truth. We propose falsifying candidate causal graphs based on whether they can explain the propagation of an outlier event. Our approach leverages a key principle: weak outliers rarely cause strong ones. While this principle has previously been used in root cause analysis to identify root causes without prior knowledge of the graph, we turn it on its head and use it to falsify candidate causal graphs whose implied outlier propagation is inconsistent with the data. To this end, we present the first statistical tests for the hypothesis that a candidate graph is the true causal graph, and show they have false positive control, power guarantees against incorrect causal graphs, and can operate with a single outlier sample.
Amazon Science