R packages 'nonlinearICP' and 'CondIndTests'

R Code for 'nonlinearICP' and 'CondIndTests'.

CondIndTests

Code for a variety of nonlinear conditional independence tests:

• Kernel conditional independence test (Zhang et al., UAI 2011),
• Residual Prediction test (based on Shah and Buehlmann, arXiv 2015),
• Invariant environment prediction,
• Invariant target prediction,
• Invariant residual distribution test,
• Invariant conditional quantile prediction (all from Heinze-Deml et al., arXiv:1706.08576).

Installation

From CRAN

install.packages("CondIndTests")

From Github with devtools

devtools::install_github("christinaheinze/nonlinearICP-and-CondIndTests/CondIndTests")

nonlinearICP

Code for 'nonlinear Invariant Causal Prediction' to estimate the causal parents of a given target variable from data collected in different experimental or environmental conditions, extending 'Invariant Causal Prediction' from Peters, Buehlmann and Meinshausen (2016) to nonlinear settings. For more details, see C. Heinze-Deml, J. Peters and N. Meinshausen: 'Invariant Causal Prediction for Nonlinear Models', arXiv:1706.08576.

Installation

From CRAN

install.packages("nonlinearICP")

From Github with devtools

devtools::install_github("christinaheinze/nonlinearICP-and-CondIndTests/nonlinearICP")

References

If you are using these packages, please cite C. Heinze-Deml, J. Peters and N. Meinshausen: 'Invariant Causal Prediction for Nonlinear Models', arXiv:1706.08576.