Introduction
Animals grazing on the same patch during a sampling period were assumed to be associating (gambit of the group)
Materials
I accessed the fowl database through networkrepository.org , an eponymous open-source log with a substantial amount of animal behavior data. This specific data on foraging patterns came from a published animal interaction study conducted by researchers in the Netherlands on data from a flock of 43 captive Barnacle Geese (Branta leucopsis). The study focused on genetic profiles, while the available data did contain hereditary information. The undirected datasheets for each sex group came in csv files, therefore no data wrangling was needed.
To interpret the network, I used the open-source platform Gephi. I used Microsoft Excel for supplementary statistical analysis.
Process
The datasheets for each geese sex group came in csv files, therefore no data wrangling was needed. I first imported the male geese file as undirected. My nodes were individual geese connected by edges, weighted by frequency. Unfortunately, the 43 geese were not given names, and the male and female cohorts had overlapping ID numbers. In order to add the female geese file to my Gephi workspace, I had to match the csv files and give female geese original identifiers.
When I ran statistics after importing both spreadsheets, only the male data (the first import) was outputted. In order to compare the weights between genders, I ran the female statistics in a separate workspace and combined in an Excel file. Since the networks were different sizes, I used this formula: Degree / ct nodes per group for each and then took the group average.
Average degree for male: .34
Average degree for female= .78
Based on this difference, I expected the female geese to be evenly clustered and the males to be dispersed.
The first layout I tried was the recommended tried-and-true ForceAtlas2 (Figure 1). I decided to size the nodes by degree (the largest being the most connected). Since this was about eating, it made sense to me the largest geese physically would be the most popular (although I have no data to confirm geese size). I also decided to identify clusters by coloring by modularity class.
Once the female clustering piqued my interest, I decided to try the OpenOrd layout which I read aims at unpacking clusters in undirected networks (Figure 2).
Interpretation
Based on the geese visualizations, foraging does seem to be a gendered practice. In the female (a) networks of both graphics, it is clear there is a clique indicated by the darker social class. In terms of degree size or gregariousness, the female geese do not vary. The behavior of both groups are similar, and it is clear females have a preference who they forage with. In the second graph, both cliques take up a similar proportion of space, no group seems to reign or have precedence over the other.
For the male (b) networks, clique formation is not apparent, but there is obvious delineations in size. Males seemingly don’t have a preference or pattern in foraging partners, and also range in how social, or popular they are. A fair amount of male geese are only connected to the larger group by a single foraging partner.
In further examination, I visualized the Betweenness Centrality of the networks to pinpoint the junkets or important connectors. I was expecting some junkets in both male and female networks, but in weight, the male network had a larger amount of connectors at various levels of importance.
Reflection
I think hereditary information could help solve the riddle of cliques within the female foraging network. I think it would be interesting to see if mates comingle or forage with their partner’s friends.
I wish I had more intimate knowledge of typical animal behavior structures, particularly in birds. I think witnessing bird flying formations as a child was my earliest introduction to the mystery of animal behavior I believe social network visualizations like this can unveil.
Kurvers, Ralf HJM, et al. “Contrasting context dependence of familiarity and kinship in animal social networks.” Animal Behaviour 86.5 (2013): 993-1001.