Workflow steps and the problems they solve

[yaslena]

Generating a graph

substeps

• choosing graph type
• chosing the edge list
• mapping source and targed columns
• including or excluding edge attributes (weights)

code

G = nx.from_pandas_edgelist(edges, "Source", "Target", edge_attr=True, create_using=nx.DiGraph())

considerations and case specific explanations

nx.from_pandas_edgelist(edges, This part specifies the "edges" table as the input edge-list to be used with the operation "from_pandas_edgelist".

"Source", "Target", This part marks the columns named "Source", "Target" which are contained in the table "edges" as the ones to be used as source and target in the network graph creation.

create_using=nx.DiGraph() marks the graph as a directed graph. On the one hand, this reflects the nature of the dataset which has direction-specific, assymetric edge data. The edge list thus contains information about which journals were the reviewing ones and which ones were (passively) reviewed by others. When investigating information flow, one would almost always go for a directed graph because it allows to trace the course and direction of the information.

edge_attr=True makes sure that the edge weights are considered, creating a weighted graph. The edge-list data not only contains information about which journals reviewed each other, but also how often they did so. The weight attibute holds the information about how often one journal was reviewed in another.

related questions

Where there journals in this network that were more reviewed than others?

This can be answered with the "out-degree" measurement, which is only available in directed graphs.

Russian and Japanese journals almost never review each other's articles. Which would be the most likely intermediaries between Russian and Japanese journals in the information flow that is represented in this network?

A directed graph is needed here, because the direction of information flow is important in this case. Other methods that can be used additionally: betweenness centrality and shortest path.