Causal Discovery from Subsampled Time Series Data by Constraint Optimization

By Antti Hyttinen, Sergey M. Plis, Matti Järvisalo, Frederick Eberhardt, David Danks. 2016

46 citations recorded by the source corpus.

This paper focuses on causal structure estimation from time series data in which measurements are obtained at a coarser timescale than the causal timescale of the underlying system. Previous work has shown that such subsampling can lead to significant errors about the system’s causal structure if not properly taken into account. In this paper, we first consider the search for the system timescale causal structures that correspond to a given measurement timescale structure. We provide a constraint satisfaction procedure whose computational performance is several orders of magnitude better than previous approaches. We then consider finite-sample data as input, and propose the first constraint optimization approach for recovering the system timescale causal structure. This algorithm optimally recovers from possible conflicts due to statistical errors. More generally, these advances allow for a robust and non-parametric estimation of system timescale causal structures from subsampled time series data.

Subfields: Artificial Intelligence, Computer Networks and Communications

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 Computer Networks and Communications

Citations in the corpus

Causal Discovery from Temporally Aggregated Time Series (2017, methodology): “Recently, a set of methods have been proposed to estimate the causal structure at the causal frequency from subsampled data without resorting to disaggregation techniques (Hyttinen et al., 2016; Gong et al., 2015; Plis et al., 2015a; Danks & Plis, 2013).”
A Survey of Learning Causality with Data (2018, background): “hat only a low-resolution version of the original time series is available, for example, we only observe the time series every k time steps [131]. The key assumption for the recently proposed methods [47, 63, 110] is that there exists a true timescale or causal frequency at which we can discover pro”
Advancing functional connectivity research from association to causation (2019, background): “Notably, several researchers have begun combining FC measures to constrain causal graphs of brain region interactions 2,48,49 , demonstrating that converging multimethod FC evidence can be used to constrain the hypothesis space (see Box 2).”
Markov equivalence of marginalized local independence graphs (2018, background): “Related research has been concerned with inferring the graph structure from subsampled time series, but under the assumption of no latent processes; see, for example, [9, 22].”
Identifiability and estimation of structural vector autoregressive models for subsampled and mixed-frequency time series. (2017, background): “Recently, the problem of causal discovery in subsampled time series has been studied drawing from methods in causal structure learning using graphical models [7, 8, 9, 10].”
Granger Causality: A Review and Recent Advances (2021, background): “Recently, causal discovery in subsampled time series has been studied with methods in causal structure learning using graphical models (Danks and Plis, 2013; Plis et al., 2015; Hyttinen et al., 2016).”
A Survey of Learning Causality with Data: Problems and Methods (2018, background): “The key assumption for the recently proposed methods [47, 63, 110] is that there exists a true timescale or causal frequency at which we can discover proper causal graph structures with the highest confidence [33].”
DYNOTEARS: Structure Learning from Time-Series Data (2020, unknown): “See for example Gong et al. (2015); Hyttinen et al. (2016); Plis et al. (2015); Cook et al. (2017).”