ges: Greedy Equivalence Search (GES)

In tzimiskes/causality: Examine The Causal Structure of Data

Description

GES is a score based causal discovery algorithm that outputs a pattern, a graph that encodes the markov equilevence class of a set of DAGs. GES contains score functions for continuous and discrete datasets. Mixed datasets will have to be treated treated as continuous or discretized completly. Other versions of ges support background knowledge, but this version does not.

Usage

ges(df, score = c("bic", "bdue"), penalty = 1, sample.prior = 1,
  structure.prior = 1)

Arguments

df	A data.frame with no missing values.
score	The scoring function to use. Use BIC for continuous data and BDeu for discrete.
penalty	Tuning parameter for bic score. Cannot be less than 0; less than 1 is probably a bad idea. Higher penalties will generate sparser graphs. Defaults to 1, which corresponds to standard BIC.
sample.prior	Second tuning parameter for BDeu score.
structure.prior	First tuning parameter for BDeu score.

Author(s)

Alexander Rix

References

Chickering DM. Optimal structure identification with greedy search. Journal of machine learning research. 2002;3(Nov):507-54.

Examples

library(causality)
ges(ecoli.df, "bic", penalty = 2)

tzimiskes/causality documentation built on Sept. 15, 2019, 8:41 p.m.