Closed dalhoomist closed 1 year ago
You are right, we omit node names, and use the columns' indexes (zero-indexed) instead. Hence, the node numbers in the directed_graph correspond to the column names in the order provided as data input.
Thank you for the clarification
Another question. The resultant network from running GaussianPC is a directed graph. but even though I provide a correlation matrix as part of the input, it is unweighted. Is there a way to generate a weighted graph? Or is acceptable to assign the correlation values as weights manually? Finally, is there a way to export the graph to be used as a weighted directed graph that can be edited/previewed/analyzed in cytoscape?
The package does not support generating a weighted graph. It only returns the estimated CPDAG, similar to the R pcalg package.
The directed graph is of type DiGraph
from the networkx
package. Hence, you could add the weights that you require manually. Also, networkx
supports transformation of the graph data into a format for cytoscape, e.g., see function cytoscape_data
of networkx
(see here).
Thanks very much for the information and guidance.
Hello,
Thank you for the putting together this package. I am testing it out and it is very fast.
My question is about the data input. I tried inputting a dataframe from panda and the algorithm runs fine with the GaussianPC function. However, I get a directed_graph with no node names. The output of directed_graph.nodes() is a series of numbers equal to the number of culumns in my input. Do the column names in the input correspond directly with the node numbers in the directed_graph?
Thank you