`R/estimate-network.R`

`pc_estimate_undirected_graph.Rd`

Uses the PC-algorithm and is mostly a wrapper around `pcalg::skeleton()`

.

`pc_estimate_undirected_graph(data, alpha = 0.01)`

- data
Input numeric data that forms the basis of the underlying graph.

- alpha
Significance level threshold applied to each test to determine conditional dependence for if an edge exists.

A `pcAlgo`

object that contains the DAG skeleton, aka undirected graph.

This function estimates the "skeleton of a DAG", meaning a graph without
arrowheads, aka an undirected graph.
The default estimation method used is the "PC-stable" method, which estimates
the *order-independent* skeleton of the DAG, meaning the order of the
variables given does not impact the results (older versions of the algorithm
were order-dependent). The method also assumes no latent variables.

An edge is determined by testing for conditional dependence between two
nodes based on the `pcalg::gaussCItest()`

. Conditional *independence* exists
when the nodes have zero partial correlation determined from a p-value based
hypothesis test against the correlation matrix of the data from the nodes.
The estimated edges exists between the *start* and *end* nodes when the
*start* and *end* variables are conditionally dependent given the subset of
remaining variables.

The help documentation of `pcalg::skeleton()`

has more details.