Animal Social Networks Research Articles

Permutation tests for hypothesis testing with animal social network data: Problems and potential solutions.

Permutation tests are widely used to test null hypotheses with animal social network data, but suffer from high rates of type I and II error when the permutations do not properly simulate the intended null hypothesis.Two common types of permutations each have limitations. Pre‐network (or datastream) permutations can be used to control ‘nuisance effects’ like spatial, temporal or sampling biases, but only when the null hypothesis assumes random social structure. Node (or node‐label) permutation tests can test null hypotheses that include nonrandom social structure, but only when nuisance effects do not shape the observed network.We demonstrate one possible solution addressing these limitations: using pre‐network permutations to adjust the values for each node or edge before conducting a node permutation test. We conduct a range of simulations to estimate error rates caused by confounding effects of social or non‐social structure in the raw data.Regressions on simulated datasets suggest that this ‘double permutation’ approach is less likely to produce elevated error rates relative to using only node permutations, pre‐network permutations or node permutations with simple covariates, which all exhibit elevated type I errors under at least one set of simulated conditions. For example, in scenarios where type I error rates from pre‐network permutation tests exceed 30%, the error rates from double permutation remain at 5%.The double permutation procedure provides one potential solution to issues arising from elevated type I and type II error rates when testing null hypotheses with social network data. We also discuss alternative approaches that can provide robust inference, including fitting mixed effects models, restricted node permutations, testing multiple null hypotheses and splitting large datasets to generate replicated networks. Finally, we highlight ways that uncertainty can be explicitly considered and carried through the analysis.

Analyzing Contacts and Behavior from High Frequency Tracking Data Using the wildlifeDI R Package

Inter‐individual interactions are one of the key factors driving patterns of wildlife movement; however, methods for capturing and analyzing inter‐individual interactions from wildlife tracking data remain limited. Extracting contacts from wildlife tracking data is a challenge owing to the complex spatial and temporal patterns and the volume of tracking data sets. Knowledge of the time and location of contacts are crucial to understanding the spatiotemporal patterns of contacts and how they relate to the environment, individual behavior, and social structure. In this article we introduce a new suite of functions in the wildlifeDI R package for automating contact analysis, summaries, and outputs (e.g., visualizations) from studies tracking many individuals simultaneously, building upon the existing methods for studying interactive behavior between dyads already present within the package. The package has applications to study contact and interaction for the study of animal behavior, social networks, and disease transmission. We demonstrate two applications of contact analysis using the wildlifeDI package: female white‐tailed deer (Odocoileus virginianus) contacts and contacts between hunters and male white‐tailed deer. The wildlifeDI package represents a new set of advanced, reproducible analyses to identify and study contacts and interactions in wildlife tracking studies. We designed the analyses and outputs to integrate into existing R analysis workflows to facilitate adoption of the package into a wide variety of wildlife tracking studies.

Dimension-reduction of dynamics on real-world networks with symmetry

We derive explicit formulae to quantify the Markov chain state-space compression, or lumping, that can be achieved in a broad range of dynamical processes on real-world networks, including models of epidemics and voting behaviour, by exploiting redundancies due to symmetries. These formulae are applied in a large-scale study of such symmetry-induced lumping in real-world networks, from which we identify specific networks for which lumping enables exact analysis that could not have been done on the full state-space. For most networks, lumping gives a state-space compression ratio of up to 10 7 , but the largest compression ratio identified is nearly 10 12 . Many of the highest compression ratios occur in animal social networks. We also present examples of types of symmetry found in real-world networks that have not been previously reported.

Influence of number of individuals and observations per individual on a model of community structure.

Social network analysis is increasingly applied to understand animal groups. However, it is rarely feasible to observe every interaction among all individuals in natural populations. Studies have assessed how missing information affects estimates of individual network positions, but less attention has been paid to metrics that characterize overall network structure such as modularity, clustering coefficient, and density. In cases such as groups displaying fission-fusion dynamics, where subgroups break apart and rejoin in changing conformations, missing information may affect estimates of global network structure differently than in groups with distinctly separated communities due to the influence single individuals can have on the connectivity of the network. Using a bat maternity group showing fission-fusion dynamics, we quantify the effect of missing data on global network measures including community detection. In our system, estimating the number of communities was less reliable than detecting community structure. Further, reliably assorting individual bats into communities required fewer individuals and fewer observations per individual than to estimate the number of communities. Specifically, our metrics of global network structure (i.e., graph density, clustering coefficient, Rcom) approached the ‘real’ values with increasing numbers of observations per individual and, as the number of individuals included increased, the variance in these estimates decreased. Similar to previous studies, we recommend that more observations per individual should be prioritized over including more individuals when resources are limited. We recommend caution when making conclusions about animal social networks when a substantial number of individuals or observations are missing, and when possible, suggest subsampling large datasets to observe how estimates are influenced by sampling intensity. Our study serves as an example of the reliability, or lack thereof, of global network measures with missing information, but further work is needed to determine how estimates will vary with different data collection methods, network structures, and sampling periods.

Age and sex influence social interactions, but not associations, within a killer whale pod.

Social structure is a fundamental aspect of animal populations. In order to understand the function and evolution of animal societies, it is important to quantify how individual attributes, such as age and sex, shape social relationships. Detecting these influences in wild populations under natural conditions can be challenging, especially when social interactions are difficult to observe and broad-scale measures of association are used as a proxy. In this study, we use unoccupied aerial systems to observe association, synchronous surfacing, and physical contact within a pod of southern resident killer whales (Orcinus orca). We show that interactions do not occur randomly between associated individuals, and that interaction types are not interchangeable. While age and sex did not detectably influence association network structure, both interaction networks showed significant social homophily by age and sex, and centrality within the contact network was higher among females and young individuals. These results suggest killer whales exhibit interesting parallels in social bond formation and social life histories with primates and other terrestrial social mammals, and demonstrate how important patterns can be missed when using associations as a proxy for interactions in animal social network studies.

The scaling of social interactions across animal species

Social animals self-organise to create groups to increase protection against predators and productivity. One-to-one interactions are the building blocks of these emergent social structures and may correspond to friendship, grooming, communication, among other social relations. These structures should be robust to failures and provide efficient communication to compensate the costs of forming and maintaining the social contacts but the specific purpose of each social interaction regulates the evolution of the respective social networks. We collate 611 animal social networks and show that the number of social contacts E scales with group size N as a super-linear power-law E=CN^beta for various species of animals, including humans, other mammals and non-mammals. We identify that the power-law exponent beta varies according to the social function of the interactions as beta = 1+a/4, with a approx {1,2,3,4}. By fitting a multi-layer model to our data, we observe that the cost to cross social groups also varies according to social function. Relatively low costs are observed for physical contact, grooming and group membership which lead to small groups with high and constant social clustering. Offline friendship has similar patterns while online friendship shows weak social structures. The intermediate case of spatial proximity (with beta =1.5 and clustering dependency on network size quantitatively similar to friendship) suggests that proximity interactions may be as relevant for the spread of infectious diseases as for social processes like friendship.

Animal social networks: an introduction for complex systems scientists

AbstractMany animals live in societies where individuals frequently interact socially with each other. The social structures of these systems can be studied in depth by means of network analysis. A large number of studies on animal social networks in many species have in recent years been carried out in the biological research field of animal behaviour and have provided new insights into behaviour, ecology and social evolution. This line of research is currently not so well connected to the field of complex systems as could be expected. The purpose of this article is to provide an introduction to animal social networks for complex systems scientists and highlight areas of synergy. We believe that an increased integration of animal social networks with the interdisciplinary field of complex systems, and networks would be beneficial for various reasons. Increased collaboration between researchers in this field and biologists studying animal social systems could be valuable in solving challenges that are of importance to animal social network research. Furthermore, animal social networks provide the opportunity to investigate hypotheses about complex systems across a range of natural real-world social systems. In this article, we describe what animal social networks are and main research themes where they are studied; we give an overview of the methods commonly used to study animal social networks; we highlight challenges in the study of animal social networks where complex systems expertise may be particularly valuable; and we consider aspects of animal social networks that may be of particular interest to complex systems researchers. We hope that this will help to facilitate further interdisciplinary collaborations involving animal social networks and further integration of these networks into the field of complex systems.

Nonparametric analysis of inter‐individual relations using an attention‐based neural network

Abstract Social network analysis, which has been widely adopted in animal studies over the past decade, enables the revelation of global characteristic patterns of animal social systems from pairwise inter‐individual relations. Animal social networks are typically drawn based on the geometric proximity and/or frequency of social behaviours (e.g. grooming), but the appropriate metric for inter‐individual relationship is not clear, especially when prior knowledge on the species/data is limited. In this study, researchers explored a nonparametric analysis of inter‐individual relations using a neural network with the attention mechanism, which plays a central role in natural language processing. The high interpretability of the attention mechanism and flexibility of the entire neural network allow for automatic detection of inter‐individual relations included in the raw data, without requiring prior knowledge/assumptions about what modes/types of relations are included in the data. For these case studies, three‐dimensional location data collected from simulated agents and real Japanese macaques were analysed. The proposed method successfully recovered the latent relations behind the simulated data and discovered female‐oriented relations in the real data, which are in accordance with the previous generalizations about the macaque social structure. The proposed method does not exploit any behavioural patterns that are particular to Japanese macaques, and researchers can use it for location data of other animals. The flexibility of the neural network would also allow for its application to a wide variety of data with interacting components, such as vocal communication. ​

Animal social networks: Towards an integrative framework embedding social interactions, space and time

Animal social networks: Towards an integrative framework embedding social interactions, space and time

Identifying Central Nodes in Directed and Weighted Networks

An issue of critical interest in complex network analysis is the identification of key players or important nodes. Centrality measures quantify the notion of importance and hence provide a mechanism to rank nodes within a network. Several centrality measures have been proposed for un-weighted, un-directed networks but applying or modifying them for networks in which edges are weighted and directed is challenging. Existing centrality measures for weighted, directed networks are by and large domain-specific. Depending upon the application, these measures prefer either the incoming or the outgoing links of a node to measure its importance. In this paper, we introduce a new centrality measure, Affinity Centrality, that leverages both weighted in-degrees as well as out-degrees of a node’s local neighborhood. A tuning parameter permits the user to give preference to a node’s neighbors in either incoming or outgoing direction. To evaluate the effectiveness of the proposed measure, we use three types of real-world networks - migration, trade, and animal social networks. Experimental results on these weighted, directed networks demonstrate that our centrality measure can rank nodes in consonance to the ground truth much better than the other established measures

Observing the unwatchable: Integrating automated sensing, naturalistic observations and animal social network analysis in the age of big data.

In the 4.5 decades since Altmann (1974) published her seminal paper on the methods for the observational study of behaviour, automated detection and analysis of social interaction networks have fundamentally transformed the ways that ecologists study social behaviour. Methodological developments for collecting data remotely on social behaviour involve indirect inference of associations, direct recordings of interactions and machine vision. These recent technological advances are improving the scale and resolution with which we can dissect interactions among animals. They are also revealing new intricacies of animal social interactions at spatial and temporal resolutions as well as in ecological contexts that have been hidden from humans, making the unwatchable seeable. We first outline how these technological applications are permitting researchers to collect exquisitely detailed information with little observer bias. We further recognize new emerging challenges from these new reality-mining approaches. While technological advances in automating data collection and its analysis are moving at an unprecedented rate, we urge ecologists to thoughtfully combine these new tools with classic behavioural and ecological monitoring methods to place our understanding of animal social networks within fundamental biological contexts.

Common datastream permutations of animal social network data are not appropriate for hypothesis testing using regression models.

1. Social network methods have become a key tool for describing, modelling, and testing hypotheses about the social structures of animals. However, due to the non-independence of network data and the presence of confounds, specialized statistical techniques are often needed to test hypotheses in these networks. Datastream permutations, originally developed to test the null hypothesis of random social structure, have become a popular tool for testing a wide array of null hypotheses in animal social networks. In particular, they have been used to test whether exogenous factors are related to network structure by interfacing these permutations with regression models. 2. Here, we show that these datastream permutations typically do not represent the null hypothesis of interest to researchers interfacing animal social network analysis with regression modelling, and use simulations to demonstrate the potential pitfalls of using this methodology. 3. Our simulations show that, if used to indicate whether a relationship exists between network structure and a covariate, datastream permutations can result in extremely high type I error rates, in some cases approaching 50%. In the same set of simulations, traditional node-label permutations produced appropriate type I error rates (~ 5%). 4. Our analysis shows that datastream permutations do not represent the appropriate null hypothesis for these analyses. We suggest that potential alternatives to this procedure may be found in regarding the problems of non-independence of network data and unreliability of observations separately. If biases introduced during data collection can be corrected, either prior to model fitting or within the model itself, node-label permutations then serve as a useful test for interfacing animal social network analysis with regression modelling.

The performance of permutations and exponential random graph models when analyzing animal networks

Abstract Social network analysis is a suite of approaches for exploring relational data. Two approaches commonly used to analyze animal social network data are permutation-based tests of significance and exponential random graph models. However, the performance of these approaches when analyzing different types of network data has not been simultaneously evaluated. Here we test both approaches to determine their performance when analyzing a range of biologically realistic simulated animal social networks. We examined the false positive and false negative error rate of an effect of a two-level explanatory variable (e.g., sex) on the number and combined strength of an individual’s network connections. We measured error rates for two types of simulated data collection methods in a range of network structures, and with/without a confounding effect and missing observations. Both methods performed consistently well in networks of dyadic interactions, and worse on networks constructed using observations of individuals in groups. Exponential random graph models had a marginally lower rate of false positives than permutations in most cases. Phenotypic assortativity had a large influence on the false positive rate, and a smaller effect on the false negative rate for both methods in all network types. Aspects of within- and between-group network structure influenced error rates, but not to the same extent. In "grouping event-based" networks, increased sampling effort marginally decreased rates of false negatives, but increased rates of false positives for both analysis methods. These results provide guidelines for biologists analyzing and interpreting their own network data using these methods.

The problem and promise of scale in multilayer animal socialnetworks.

Scale remains a foundational concept in ecology. Spatial scale, for instance, has become a central consideration in the way we understand landscape ecology and animal space use. Meanwhile, scale-dependent social processes can range from fine-scale interactions to co-occurrence and overlapping home ranges. Furthermore, sociality can vary within and across seasons. Multilayer networks promise the explicit integration of the social, spatial, and temporal contexts. Given the complex interplay of sociality and animal space use in heterogeneous landscapes, there remains an important gap in our understanding of the influence of scale on animal social networks. Using an empirical case study, we discuss ways of considering social, spatial, and temporal scale in the context of multilayer caribou social networks. Effective integration of social and spatial processes, including biologically meaningful scales, within the context of animal social networks is an emerging area of research. We incorporate perspectives that link the social environment to spatial processes across scales in a multilayer context.

Is less more? A commentary on the practice of ‘metric hacking’ in animal social network analysis

The use of social network analysis to quantify animal social relationships has increased exponentially over the last two decades. A popular aspect of social network analysis is the use of individually based network metrics. Despite the diversity of social network metrics that exist and the large number of studies that generate network metrics, little guidance exists on the number and type of metrics that should be analysed in a single study. Here, we comment on the ‘hypothesize after results are known’ (HARKing) phenomenon in the context of social network analysis, a practice that we term ‘metric hacking’ and define as the use of statistical criteria to select which metrics to use rather than a priori choice based on a research hypothesis. We identify three situations where metric hacking can occur in studies quantifying social network metrics: (1) covariance among network metrics as explanatory variables in the same model; (2) covariance among network metrics as response variables in multiple models; and (3) covariance between response and explanatory variables in the same model. We outline several quantitative and qualitative issues associated with metric hacking, provide alternative options and guidance on the appropriate use of multiple network metrics to avoid metric hacking. By increasing awareness of the use of multiple social network metrics, we hope to encourage better practice for the selection and use of social network metrics in animal social network analysis.

Infected or informed? Social structure and the simultaneous transmission of information and infectious disease

Social interactions present opportunities for both information and infection to spread through populations. Social learning is often proposed as a key benefit of sociality, while infectious disease spread are proposed as a major cost. Multiple empirical and theoretical studies have demonstrated the importance of social structure for the transmission of either information or harmful pathogens and parasites, but rarely in combination. We provide an overview of relevant empirical studies, discuss differences in the transmission processes of infection and information, and review how these processes have been modelled. Finally, we highlight ways in which animal social network structure and dynamics might mediate the tradeoff between the sharing of information and infection. We reveal how modular social network structures can promote the spread of information and mitigate against the spread of infection relative to other network structures. We discuss how the maintenance of long‐term social bonds, clustering of social contacts in time, and adaptive plasticity in behavioural interactions, all play important roles in influencing the transmission of information and infection. We provide novel hypotheses and suggest new directions for research that quantifies the transmission of information and infection simultaneously across different network structures to help tease apart their influence on the evolution of social behaviour.

A multilevel statistical toolkit to study animal social networks: the Animal Network Toolkit Software (ANTs) R package

The possible role played by individual attributes, sociodemographic characteristics and/or ecological pressures in the interaction between animals and the development of social relationships between them is of great interest in animal ecology and evolutionary biology. Social Network Analysis is an ideal tool to study these types of questions. The Animal Network Toolkit Software (ANTs) R package was specifically developed to provide all the different social network analysis techniques currently used in the study of animal social networks. This global package enables users to (1) compute global, polyadic and nodal network measures; (2) perform data randomisation: data stream and network (node and link) permutations; (3) perform statistical permutation tests for static or temporal network analyses, and (4) visualise networks. ANTs allows researchers to perform multilevel network analyses ranging from individual network measures to interaction patterns and the analysis of the overall network structure, and carry out static or temporal network analyses without switching between different R packages, thus making a substantial contribution to advances in the study of animal behaviour. ANTs outperforms existing R packages for the computation speed of network measures and permutations.

How consistently do personality attributes relate to an individual’s position within a social network: a comparison across groups of captive meerkats

The social network structure of a group arises from the patterns of association of its constituent members, which in turn originates from behaviors and preferences of those individuals. Consequently, an individual’s position in their social environment is commonly related to their own attributes. However, most studies that report such relationships are limited to single or very small numbers of groups. I consider one set of attributes suggested to influence network structure, namely personality type, and observed 15 groups of captive meerkats Suricata suricatta. Thus, I could assess whether the relationship between individual attributes and patterns of association and network position persisted across groups. Principal component analysis of behavior, within the 15 groups, revealed two personality dimensions: friendliness and aggressiveness. For a subset of five groups, only friendliness was consistent over 1 year. I found little evidence that they were universally good predictors of social network structure or individual positions within networks. Individuals with high friendliness scores were more central in networks of foraging competitions. There was no evidence that meerkats preferentially associated with or avoid others based on each of their personality scores. Alternatively, and contrary to much-published work, it may not be possible to generalize relationships between individual attributes and network position or overall network structure. This may be because social environments emerge from a complex interplay between individual attributes and social interactions, and thus, the role of these elements along with the physical environment in which they live in shaping network structures and personality is challenging to disentangle. Animal personalities and animal social network structures are increasingly recognized as significant components in animal behavior and welfare. Here, I explore personality in the context of social networks in different captive groups of meerkats. These types of studies are uncommon probably because collecting detailed interaction data from multiple known individuals in multiple groups is time-consuming and because groups are often highly variable in composition and ecological context. When I did consider the social networks and individual attributes of members of 15 groups, I found little evidence that an individual’s personality scores consistently explained either their network position or patterns of association across all groups. Individuals in the wild, with different personality profiles, may adjust their network assortment to balance their costs and benefits and shift their environmental pressures. In a captive environment, such characteristics can vary and therefore social associations may be expressed differently from their wild counterparts.

Microbial transmission in animal social networks and the social microbiome.

Host-associated microbiomes play an increasingly appreciated role in animal metabolism, immunity and health. The microbes in turn depend on their host for resources and can be transmitted across the host's social network. In this Perspective, we describe how animal social interactions and networks may provide channels for microbial transmission. We propose the 'social microbiome' as the microbial metacommunity of an animal social group. We then consider the various social and environmental forces that are likely to influence the social microbiome at multiple scales, including at the individual level, within social groups, between groups, within populations and species, and finally between species. Through our comprehensive discussion of the ways in which sociobiological and ecological factors may affect microbial transmission, we outline new research directions for the field.

Calculating effect sizes in animal social network analysis

Abstract Because of the nature of social interaction or association data, when testing hypotheses using social network data it is common for network studies to rely on permutations to control for confounding variables, and to not also control for them in the fitted statistical model. This can be a problem because it does not adjust for any bias in effect sizes generated by these confounding effects, and thus the effect sizes are not informative in the presence of confounding variables. We implemented two network simulation examples and analysed an empirical dataset to demonstrate how relying solely on permutations to control for confounding variables can result in highly biased effect size estimates of animal social preferences that are uninformative when quantifying differences in behaviour. Using these simulations, we show that this can sometimes even lead to effect sizes that have the wrong sign and are thus the effect size is not biologically interpretable. We demonstrate how this problem can be addressed by controlling for confounding variables in the statistical dyadic or nodal model. We recommend this approach should be adopted as standard practice in the statistical analysis of animal social network data.