Causal Discovery from Changes

By Jin Tian, Judea Pearl. 2001

182 citations recorded by the source corpus.

We propose a new method of discovering causal structures, based on the detection of local, spontaneous changes in the underlying data-generating model. We analyze the classes of structures that axe equivalent relative to a stream of distributions produced by local changes, and devise algorithms that output graphical representations of these equivalence classes. We present experimental results, using simulated data, and examine the errors associated with detection of changes and recovery of structures.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Information Systems, Causal Inference & Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Information Systems

In Causal Inference & Discovery

Citations in the corpus

An Analysis of the Adaptation Speed of Causal Models (2020, background): “Tian and Pearl (2001) first studied this setting, proposing a constraint based method to infer the interventional equivalence class from a sequence of interventions.”
Toward Causal Representation Learning (2021, background): “Hence, distribution shifts, such as observing a system in different “environments/contexts,” can signiﬁcantly help to identify causal structure [188], [251].”
Towards Causal Representation Learning (2021, background): “[237, 188, 229, 113, 174, 187, 139, 17, 246, 277, 175, 123, 186, 176, 36, 82, 161], learning causal relations frequently requires collecting data from multiple environments, or the ability to perform interventions [251].”
Algorithmic recourse under imperfect causal knowledge: a probabilistic approach (2020, background): “Access to instrumental variables or experimental data may help further relax these assumptions [3, 8, 40].”
Domain Adaptation by Using Causal Inference to Predict Invariant Conditional Distributions (2017, background): “…(“surgical”) interventions modeled by the do-operator of Pearl [2009], but can also be other types of interventions such as mechanism changes [Tian and Pearl, 2001], soft interventions [Markowetz et al., 2005], fat-hand interventions [Eaton and Murphy, 2007], activity interventions [Mooij…”
Multi-Source Domain Adaptation: A Causal View (2015, background): “Even if this is not the case, multiple source domains with different data distributions may allow one to identify the causal structure, since the causal knowledge can be seen from the change in data distributions; see e.g., (Tian and Pearl 2001).”
Characterization and greedy learning of interventional Markov equivalence classes of directed acyclic graphs (2011, background): “Theorem 10 agrees with Theorem 3 of Tian and Pearl (2001) for single-variable interventions.”
Causal Structure Learning (2017, methodology): “Other excluded methods that make use of interventional data include Cooper and Yoo (1999); Tian and Pearl (2001) and Eaton and Murphy (2007), where the latter does not require knowledge of the precise location of interventions in a similar spirit to Rothenh¨ausler et al. (2015).”
Anchor regression: Heterogeneous data meet causality (2018, background): “Shift interventions can be seen as a special case of a “parametric”, “imperfect” or “dependent” intervention or a “mechanism change” [Eberhardt and Scheines, 2007, Korb et al., 2004, Tian and Pearl, 2001].”
A survey of algorithmic recourse: definitions, formulations, solutions, and prospects (2020, background): “ity, they do sounderspecificationofacausallysufficientgraph. Futureresearch in this direction may explore less strict settings, perhaps accounting for hidden confounders, or partially observed graphs [9, 44, 168], further adding to the uncertainy of recourse recommendations. Alternatively, sources o”
Probabilistic Soft Interventions in Conditional Gaussian Networks (2005, methodology): “Tian and Pearl (2001) give an example for discrete networks, which can easily be extended to Gaussian networks by p(βγ , σ2γ | do(Yγ = k)) = d(β(0)γ − k) ∏ i∈pa(γ) d(β(i)γ ) · d(σ2γ).”
Restructuring Dynamic Causal Systems in Equilibrium (2005, background): “There has been much success on the topic of causal discovery in the past decade in Artificial Intelligence [Spirtes et al., 2000; Verma and Pearl, 1991; Heckerman et al., 1999; Tian and Pearl, 2001], building on structural-equation modelling techniques originating in early econometrics [cf., Simon, ”
Causal Dantzig: Fast inference in linear structural equation models with hidden variables under additive interventions (2017, background): “…1989, Robins et al., 2000, Pearl, 2009], one possibility is to characterize the Markov equivalence class of graphs under the assumption of acyclicity and usually faithfulness [Verma and Pearl, 1991, Andersson et al., 1997, Tian and Pearl, 2001, Hauser and B¨uhlmann, 2012, Chickering, 2002].”
Learning Causal Structures Using Regression Invariance (2017, methodology): “In [35], the authors analyzed the classes of structures that are equivalent relative to a stream of distributions and presented algorithms that output graphical representations of these equivalence classes.”
Repair of Partly Misspecified Causal Diagrams. (2017, background): “A related notion, where an expert provides full parent sets for a restricted subset of variables, is discussed in reference 29.”
Causal Discovery from Soft Interventions with Unknown Targets: Characterization and Learning (2020, background): “This operation is also known as a mechanism change [31] or a parameter change [5], and it presents in many settings a more realistic model than hard or perfect interventions, where variables are forced to a fixed value (see also [2, 3, 33, 22, Sec.”