Abstract:Inferring the effect of interventions within complex systems is a fundamental problem of statistics. A widely studied approach employs structural causal models that postulate noisy functional relations among a set of interacting variables. The underlying causal structure is then naturally represented by a directed graph whose edges indicate direct causal dependencies. In a recent line of work, additional assumptions on the causal models have been shown to render this causal graph identifiable from observational data alone. One example is the assumption of linear causal relations with equal error variances that we will take up in this work. When the graph structure is known, classical methods may be used for calculating estimates and confidence intervals for causal effects. However, in many applications, expert knowledge that provides an a priori valid causal structure is not available. Lacking alternatives, a commonly used two-step approach first learns a graph and then treats the graph as known in inference. This, however, yields confidence intervals that are overly optimistic and fail to account for the data-driven model choice. We argue that to draw reliable conclusions, it is necessary to incorporate the remaining uncertainty about the underlying causal structure in confidence statements about causal effects. To address this issue, we present a framework based on test inversion that allows us to give confidence regions for total causal effects that capture both sources of uncertainty: causal structure and numerical size of nonzero effects.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to make reliable inferences about causal effects in linear causal models when the structure is uncertain. Specifically, the author focuses on how to construct confidence intervals that can take into account both the uncertainty of the causal structure and the uncertainty of the magnitude of causal effects in the absence of prior knowledge of the causal structure. Traditional methods usually estimate the potential causal structure in the first step through causal learning algorithms and then calculate the confidence intervals of causal parameters in the second step using classical statistical inference methods. However, this method ignores the uncertainty brought about by data - driven model selection, resulting in overly optimistic confidence intervals that fail to achieve the expected coverage probability. To solve this problem, the author proposes a framework based on hypothesis - test inversion, which allows for the construction of confidence regions for the total causal effect that can capture two types of uncertainty (the causal structure and the magnitude of non - zero effects). In this way, researchers can more reliably assess the existence and strength of causal effects without prior knowledge of the causal structure. The key contribution of the paper lies in providing a new method to deal with the uncertainty of the causal structure, which is of great significance for predicting the effects of interventions in complex systems. By strictly considering structural uncertainty, this method can improve the reliability of causal inferences, especially in application settings where the exact causal structure is often difficult to determine or completely unknown.

Confidence in Causal Inference under Structure Uncertainty in Linear Causal Models with Equal Variances

Confidence in Causal Discovery with Linear Causal Models

Dual Likelihood for Causal Inference under Structure Uncertainty

Identifying Total Causal Effects in Linear Models under Partial Homoscedasticity

A cautious approach to constraint-based causal model selection

Causal Inference under Interference and Model Uncertainty

Nonlinearity, Feedback and Uniform Consistency in Causal Structural Learning

Amortized Inference for Causal Structure Learning

Learning Causal Structures Based on Divide and Conquer

Causal Inference with Non-IID Data under Model Uncertainty

Causal Discovery of Linear Non-Gaussian Causal Models with Unobserved Confounding

Causal Discovery in Linear Models with Unobserved Variables and Measurement Error

Discovering Causal Models with Optimization: Confounders, Cycles, and Instrument Validity

Partial Homoscedasticity in Causal Discovery with Linear Models

Causal inference for community-based multi-layered intervention study.

Towards Bounding Causal Effects under Markov Equivalence

Conducting Simulations in Causal Inference with Networks-Based Structural Equation Models

Causality on cross-sectional data: Stable specification search in constrained structural equation modeling

Causal Discovery in Linear Structural Causal Models with Deterministic Relations

An Approach to Causal Inference over Stochastic Networks

Statistical Approaches for Causal Inference