pathfinder: The pathfinder network
In sparsebn: Learning Sparse Bayesian Networks from High-Dimensional Data

Description Usage Format Details References Examples

Simulated data and network for the pathfinder network from the Bayesian network repository. Pathfinder is an expert system developed by Heckerman et. al (1992) [1] to assist with the diagnosis of lymph-node diseases.

1	data(pathfinder)

A list with four components:

dag An edgeList containing the pathfinder network (109 nodes, 195 edges).
data A data.frame with 109 variables and 1000 observations.
ivn A list specifying which nodes are under intervention in each observation; since this dataset is purely observational, this is just NULL. Compatible with the input to sparsebnData.
cov Covariance matrix used to generate the data.

This is a benchmark network used to test algorithms for learning Bayesian networks. The data is simulated from a Gaussian SEM assuming unit edge weights and unit variances for all nodes.

[1] Heckerman, David E., and Bharat N. Nathwani. "An evaluation of the diagnostic accuracy of Pathfinder." Computers and Biomedical Research 25.1 (1992): 56-74.

### Create a valid sparsebnData object from the simulated pathfinder data
data(pathfinder)
dat <- sparsebnData(pathfinder$data, type = "c")

### Code to reproduce this dataset by randomly generating edge weights
coefs <- runif(n = num.edges(pathfinder$dag), min = 0.5, max = 2) # coefficients
vars <- rep(1, num.nodes(pathfinder$dag))                         # variances
params <- c(coefs, vars)                                          # parameter vector
pathfinder.data <- generate_mvn_data(graph = pathfinder$dag,
                                     params = params,
                                     n = 1000)