SOCS0081--Social Networks, Assessment 2

INSTRUCTIONS:

- The assessment is due on Tues May 13th 2025, 1pm and shall be submitted via Moodle.

- Late submission results in penalties, see: https://www.ucl.ac.uk/academic-manual/chapters/chapter-4-assessment-framework-taught-programmes/section-3-module-assessment#3.12. There is no exception to late submission penalties, unless an extenuating circumstances application has been successfully made.

- Submit on Moodlea single document that includes the main body of your report and any tables and figures you may use in your report. Any R code you use to produce your results should be given in an appendix. If you use some other software than R, do include any code you have used with details of the software you used.

- Word limit is 1,500. This excludes tables, figures, table and figure legends/captions, references and the appendix, but includes footnotes and endnotes. Exceeding this limit will result in penalties.

- At the end of your essay include the number of words of your report, excluding the tables, figures, table and figure legends, references (if you used any), and the appendix with R code.

- If you use any reference in your report, list full bibliographic details at the end of your report. Any referencing style. (ASA, APA, Harvard, Chicago etc.) is fine, provided that the style. is used consistently.

- The coursework will be assessed against the criteria set in the UCL UG-ESSAY GRADING SCHEME, a pdf of which could be found on the programme handbooks. In addition to those general guidelines, further specific factors will affect the marks: correctness of the solutions and interpretations of results, clarity of arguments, rigour in presenting and analysing the network, creativity in your approach, and the ability to demonstrate that key concepts treated in the course are understood well.

- This is an assessed piece of coursework for the SOCS0081 module; collaboration and/or discussion of the assessment with anyone is strictly prohibited. The rules for plagiarism apply and any cases of suspected plagiarism of published work or the work of classmates will be taken seriously.

- This assessment is a Category 2 assessment which means that GenAI tools can only be used in an assistive role. In particular, you can use tools like ChatGPT or CoPilot to assist with R coding and proof-reading, but the data collection, analysis and write-up should be your own.

- For more tips please look at the SNA Cookbook file on Moodle.

- Please remember to define and explain technical concepts, measurements etc to demonstrate that you understand what you are doing.

In the 1st summative assessment, you built your own network. In this second assessment, you will continue using your network from assessment 1. You may build a new network for this exercise too (bearing in mind that this may require extra work for you). If you decide to build a new network, adhere to the constraints given in assessment 1 on how your network should look like (e.g. the network should have at least 25 nodes, the network should be original etc.) Consult Assessment 1 on Moodle for the requirements for your network. Note that you may need to ignore weights, directions, or signs of edges for some of the algorithms you’ll use below. If this turns out to be the case, mention briefly that the algorithm you use ignores (or you choose to ignore) some characteristics of the edges. Some algorithms you’ll use below may fail to converge. If this happens, report the case, modify the algorithm, the statistical model, or your network until you get a solution.

You will write 1,500 words report. Your report should discuss the items given below. Structure your report in four parts corresponding to the four groups of items below. Part C weighs 34%, the other parts weigh 22% each.

A: Assortativity and communities

First describe briefly your network (i.e. what are nodes and edges) and how you constructed the network (i.e. how you collected the data). Also provide a plot of your network. The purpose is to remind us of your network. If you chose to build a new network for this assessment, you will need give more details here.

1. Divide your network into two or three mutually exclusive subgroups. There may already be natural subgroups in your network (e.g. defensive versus offensive football players, actors from different teams, war lords from different clans, gender, students from different schools or countries etc.). If this is the case, use these natural divisions. If your network does not have such natural subgroups, impose an artificial division yourself and justify your division. Calculate a modularity statistic using these subgroups and interpret the results. What do the results imply for assortativity in your network?

2. Now study assortativity with respect to a continuous variable, except degree. This continuous variable could be any variable (e.g. age, income, or some other network variable of a node) except degree. Interpret the results. What do your results imply for the level of assortativity in your network with respect to the continuous variable you study?

3. Ignore now the division you imposed/studied in A-1. Run a community detection algorithm to detect hidden communities. Settle on a final community structure. Compare the communities you find here with those in A-1. Interpret the results of your community detection algorithm.

B: Small-world and scale-free networks

Discuss briefly what a scale-free network is and what the small-world phenomenon means. Report the degree distribution in your network and some measures of distance between the nodes in your network. Discuss if your network looks like a scale-free network and exhibits small-world characteristics (Your network will likely be rather small to discuss these properties which apply to very large networks. But imagine you expand your network by adding many more nodes or by collecting additional data from many other similar networks. Would you expect to see scale-free or small-world network characteristics?). Briefly discuss the mechanism that may or may not result in a scale-free and a small-world network in your case.

C: Exponential random graph modelling

Discuss briefly what Exponential Random Graph Modelling (ERGM) can tell us about your network that other approaches we treated in this class cannot tell. Formulate at least three hypotheses that you can test using an ERGM. Test these hypotheses by fitting an ERGM. Interpret the results. Carry out a simulation analysis to assess the goodness-of-fit of your ERGM and interpret the results. [NB: not all ERGMs converge. If non-convergence (R failing to find reasonable solutions) occurs in your case, report this, and try different specifications (e.g. adding geometrically weighted terms, varying the decay parameter of these terms, removing terms, adding alternative terms) until you achieve convergence. If you cannot achieve convergence by this way, try modifying your network by e.g. adding a few new links or removing certain links, expanding your network by adding a few new nodes etc. If nothing works after all these steps, contact the teaching team.]

D: Self-reflection on Social Network Analysis

Based on your experience of the analysis of your network in assessment 1 and assessment 2, and comparing it with other social science approaches you have seen during your study, in ~350 words, evaluate critically SNA discussing its key features and advantages as well as potential drawbacks.