search

ysamwang / highDNG

high dimensional graph discovery
4 stars 0 forks source link

Prior Knowledge Given to findGraphMulti #1

Open gabriel-ruiz opened 4 years ago

gabriel-ruiz commented 4 years ago

Hi,

Regarding Section 4.3 Pre-selection of neighborhoods in [1] and the corresponding example (I think) in [...]/tests/highD_preSelect.R [2]: what is the format of skel in findGraphMulti(Y, maxInDegree = maxInDegree, cutOffScaling = cs, B = skel, degree = deg, verbose = T))?

For example, if I run neighborhood lasso regression to get the support for an undirected graph, should skel be the corresponding p by p matrix with 1's and 0's?

[1] Y. S. Wang and M. Drton. "High-dimensional causal discovery under non-Gaussianity." Biometrika (2020). [2] https://github.com/ysamwang/highDNG/blob/master/highDlingam/tests/highD_preSelect.R

Thank you for posting this code. It has worked great so far.

ysamwang commented 4 years ago

Yes! It should be 1's (indicating possible edge) and 0's (indicating no edge) with all diagonal elements also being 0

gabriel-ruiz commented 4 years ago

Awesome, thank you!