Data Generating Process to Evaluate Causal Discovery Techniques for Time Series Data

Causal knowledge is highly valuable but difficult to discover and hidden from view. This creates a challenge for evaluating causal discovery algorithms. causaLens have developed a flexible framework for generating synthetic data which enables evaluation of time series causal discovery methods. We use our framework to benchmark prominent methods.

Download Research Paper

causaLens’ Andrew Lawrence, Marcus Kaiser, Rui Sampaio, and Maksim Sipos introduce a framework for evaluating causal discovery techniques for time-series data.

Data Generating Process to Evaluate Causal Discovery Techniques for Time Series Data

Related Research Papers

Ranking Policy Decisions

Unsuitability of NOTEARS for Causal Graph Discovery

An Overview of the Methodologies of Causal Discovery

Explanations for Occluded Images

On Testing for Discrimination Using Causal Models