Efficient selection of undirected graphical models for high-dimensional datasets

The gRapHD package provides functions for efficient selection of undirected graphical models (Markov networks) for high-dimensional datasets. The model variables may be discrete, continuous or both. A simple example using the Iris dataset is as follows

data(iris)
gF <- minForest(iris)
gD <- stepw(gF, data=iris)
plot(gD)

The minForest function finds the minimum BIC forest for the dataframe Iris. The stepw function finds the decomposable graphical model with minimum BIC, using forward selection starting out from gF. The plot.gRapHD function displays the graph of the model. Both gF and gD are gRapHD objects, which represent graphical models as lists of edges and vertices. The package also contains a variety of utility functions for working with gRapHD objects that are useful in high-dimensional modelling.

References

Gabriel C. G. de Abreu, Rodrigo Labouriau, David Edwards, (2009). High-dimensional Graphical Model Search with gRapHD R Package. arXiv:0909.1234v2.

[Package Contents]