graphID.genericID: Determine generic identifiability of a mixed graph.

View source: R/graphID.R

graphID.genericIDR Documentation

Determine generic identifiability of a mixed graph.

Description

If the directed part of input graph is cyclic then will check for generic identifiability using the half-trek criterion. Otherwise will use the a slightly stronger version of the half-trek criterion using ancestor decompositions.

Usage

graphID.genericID(L, O)

Arguments

L

Adjacency matrix for the directed part of the path diagram/mixed graph; an edge pointing from i to j is encoded as L[i,j]=1 and the lack of an edge between i and j is encoded as L[i,j]=0. There should be no directed self loops, i.e. no i such that L[i,i]=1.

O

Adjacency matrix for the bidirected part of the path diagram/mixed graph. Edges are encoded as for the L parameter. Again there should be no self loops. Also this matrix will be coerced to be symmetric so it is only necessary to specify an edge once, i.e. if O[i,j]=1 you may, but are not required to, also have O[j,i]=1.

Value

The vector of nodes that could be determined to be generically identifiable.

References

Foygel, R., Draisma, J., and Drton, M. (2012) Half-trek criterion for generic identifiability of linear structural equation models. Ann. Statist. 40(3): 1682-1713.

Drton, M. and Weihs, L. (2015) Generic Identifiability of Linear Structural Equation Models by Ancestor Decomposition. arXiv 1504.02992


Lucaweihs/SEMID documentation built on July 22, 2023, 7:49 a.m.