Temporal Network Analysis with R - errors running code

[acrymble]

• Temporal Network Analysis with R (tested in English only)
• Mac, running R v 4.2.2.

I tried the code exactly as posted, and started having divergent results from the tutorial in the section starting "Once we have imported this temporal data".

# Make the temporal network dynamicCollabs <- networkDynamic( thenetwork, edge.spells = PHDynamicEdges, vertex.spells = PHDynamicNodes )

Returned the following:

Edge activity in base.net was ignored Created net.obs.period to describe network Network observation period info: Number of observation spells: 1 Maximal time range observed: 1257.5 until 1325 Temporal mode: continuous Time unit: unknown Suggested time increment: NA

Which I don't think is the expected result, and suggests something went wrong.

From there, the line:

plot(dynamicCollabs)

does not return a similar looking result to the (unnumbered) figure, despite the text saying it should.

The next code:

filmstrip(dynamicCollabs, displaylabels = FALSE)

gives error message:

No coordinate information found in network, running compute.animation Error in as.matrix.network.adjacency(net, attrname = weight.attr, expand.bipartite = TRUE) : Multigraphs not currently supported in as.matrix.network.adjacency. Exiting.

All further code also returns error messages. I wasn't able to get this to resolve. It did work for me and my students this time last year, so I suspect an issue with either newer R versions or a dependence updating and breaking things.

[rivaquiroga]

I get the same errors in the Spanish version

[hawc2]

This lesson might be outdated now, and may not be easy to update.

It's odd because the first problem sounds related to the data/analysis, whereas the second problem sounds related to outdated software. So at first glance it's not clear if there's two layers of issues here to address.

@ZoeLeBlanc I see you we're a reviewer on this, do you have any thoughts on this bug?

[ZoeLeBlanc]

Hmm we would need to see the code and data to resolve this issue. Looking at the github repo, seems like this might be a similar issue statnet/ndtv#50 but I'm not seeing anything on that repo to suggest this library is fully deprecated

[jenniferisasi]

May we contact the author in case he's been working with the code lately?

[ZoeLeBlanc]

Interestingly the issue I posted above was someone trying to work through our lesson 😅 statnet/networkDynamic#14 (comment) . The person claimed that this link https://bomiklee.com/post/making-network-animations-using-ndtv-package/ helped them solve the issue by making a network list (not sure what that means). I did try running some of the lesson code and was able to get the initial plots working (though did have to wrap the plot function in the R View function).

I'm not an R person, so had ChatGPT rewrite the R code from that blogpost into Python:

import pandas as pd
import networkx as nx

nv = len(vtx)

rivlist = []

for i in range(1, 5):
    rivi = riv[riv['no'] == i]
    rivii = rivi[['stateabb1', 'stateabb2']].values
    netyr = nx.Graph()
    netyr.add_nodes_from(vtx['stateabb'])
    
    for edge in rivii:
        netyr.add_edge(edge[0], edge[1])
    
    nx.set_node_attributes(netyr, {node: i for node in netyr.nodes}, 'order')
    rivlist.append(netyr)

#Also, note that in this Python code, the vtx and riv variables should be Pandas DataFrames, corresponding to the R data.frames in the original R code.

def set_vertex_attribute(graph, attribute_name, attribute_values):
    nx.set_node_attributes(graph, attribute_values, attribute_name)

set_vertex_attribute(rivlist[0], 'vtx87', vtx87)

cols1 = []
for i in range(204):
    node = rivlist[0].nodes[i]
    ccode = rivlist[0].nodes[node]['vtx87']
    
    if ccode < 200:
        color = "red"
    elif 200 <= ccode < 400:
        color = "orange"
    elif 400 <= ccode < 630:
        color = "yellow"
    elif 630 <= ccode < 700:
        color = "green"
    else:
        color = "blue"
        
    cols1.append(color)

#In this Python code, the vtx87 variable should be a dictionary with keys as the node names and values as the corresponding attributes from the R code. For example, if in R the names(vtx87) is c("USA", "CAN") and vtx87 is c(2, 20), then in Python, the vtx87 should be {"USA": 2, "CAN": 20}.

I think contacting the author would make sense since this is a pretty chunking block of code to add. Instead of rewriting all the code, we could also ask the author to make the lesson not include multigraphs as an option too.

[anisa-hawes]

Thank you, all. I really appreciate your comments.

I'm happy to write to the author, and ask for their input to move forwards. I will update this Issue when I have further information.

[anisa-hawes]

Note: I have written to the author to explain the Issue and invite their contribution.

[alexbrey]

Thanks for reaching out. I'm not sure when one of the packages/dependencies that caused this was updated but I believe this is a problem caused by the "multiple = TRUE" attribute for the initial static network that serves as a base for the dynamic network. It should be resolvable by changing that to "multiple = FALSE"

The "No coordinate information found in network, running compute.animation" and "Edge activity in base.net was ignored Created net.obs.period to describe network Network observation period info: Number of observation spells: 1 Maximal time range observed: 1257.5 until 1325 Temporal mode: continuous Time unit: unknown Suggested time increment: NA" are just normal alerts, not errors.

[charlottejmc]

Thank you @alexbrey for getting back to us on this! Your help is very much appreciated.

@hawc2, I've opened branch Issue-2868 to make this small change to the code. I also added two info boxes to notify readers that the messages returned are normal alerts, not errors.

I was wondering whether you might have a moment to look through the lesson and decide whether it would need additional testing, or whether you're happy to merge it as is?

Thank you very much!

[jenniferisasi]

Oh! @charlottejmc please tag me too whenever Alex approves of the changes in EN so we can add the changes/edits in the ES version (that I happen to have translated myself). ¡Gracias! Thanks @alexbrey!

[hawc2]

@charlottejmc I can take a closer look.

@alexbrey that's good to hear the messages are alerts and not errors, but it sounds like @acrymble also generated a plot that looked different than the one described in the lesson?

@ZoeLeBlanc seems to have successfully generated the plot as it appears in the lesson when she reran the code though, so maybe it is working?

That's the only outstanding question I see here, otherwise seems like the changes you mentioned Charlotte should be sufficient