Ranked-Based Causal Discovery for Post-Nonlinear Models

Official implementation of rank-based PNL causal discovery methods RankG and RankS

To run the methods the following installations are required. Start R in the current directory and type

install.packages("EnvStats")
install.packages("dHSIC")

To run an example of RankG method run the following R script in the current directory

Rscript run_RankG.R

which generates PNL model with sample size 100 and runs RankG method and outputs the estimated causal order.

To run an example of RankS method run the following R script in the current directory

Rscript run_RankS.R

which generates PNL model with sample size 30 and runs RankS method and outputs the estimated causal order.