Sebastian Sosa

- Bayesian modeling is a powerful paradigm in modern statistics and machine learning. However, practitioners face significant obstacles in building bespoke models. - The landscape of Bayesian software is fragmented across programming languages and abstraction levels. Newcomers often gravitate towards high-level interfaces, like R, in order to use simple generalized linear models (GLMs) through interfaces like brms. - For niche problems, researchers must often transition to writing directly in lower-level programming languages, like Stan or JAX, which require specialist knowledge. - Furthermore, computational demands remain a significant bottleneck, often limiting the feasibility of applying Bayesian methods on large datasets and complex, high-dimensional models. - The Bayesian Inference BI is a cross-platform software distributed as a Python, R and Julia library. It provides an intuitive model-building syntax with the flexibility of low-level abstraction coding, while also providing pre-built GLM functions. Further, by facilitating hardware-accelerated GPU computation under-the-hood, BI permits high-dimensional models to be fit in a fraction of the time of comparable Stan models (up to 200-fold).

The social networks of interest to biologists, ecologists and social scientists are often multi-layered, with the same set of individuals interacting with one another in complex, multifaceted ways. Each type of interaction can be represented as one layer in a larger multiplex network. Important research questions often hinge on how ties or flows in one network layer impact ties or flows in another layer. Similar questions focus on the relationship between nodal characteristics across layers: for example, is an individual with a high out-degree in one layer more likely to have a high in-degree in another layer? These questions are effectively addressed by multiplex extensions of generative social network models, like the social relations model (SRM). Here, we present a multiplex implementation of the SRM in the STRAND R package and provide tutorials teaching end-users how to run the model on their own data. We provide worked examples of data analysis, parameter visualization and results interpretation, using datasets from experimental economics and animal behaviour. Our software package allows end-users to deploy powerful multiplex models, using only simple base-R model syntax, permitting wider use of generative network modelling approaches across disciplines.

Data collection biases are a persistent issue for studies of social networks. This issue has been particularly important in animal social network analysis (ASNA), where data are often unevenly sampled and such biases may potentially lead to incorrect inferences about animal social behaviour. Here, we address the issue by developing a Bayesian model, which not only estimates network structure but also explicitly accounts for sampling and censoring biases. Using a set of simulation experiments designed to reflect various sampling and observational biases encountered in real‐world scenarios, we systematically validate our model and evaluate its performance relative to other common ASNA methodologies. By accounting for differences in node‐level censoring (i.e. individual variation in undetected ties), our model permits the recovery of true latent social connections, even under a wide range of conditions where some key individuals are intermittently unobserved. Our model outperformed all other existing approaches and accurately captured network structure, as well as individual‐level and dyad‐level effects. Antithetically, permutation‐based and simple linear regression approaches performed the worst across many conditions. These results highlight the advantages of generative network models for ASNA, as they offer greater flexibility, robustness and adaptability to real‐world data complexities. Our findings underscore the importance of generative models that jointly estimate network structure and measurement biases typical in empirical studies of animal social behaviour.

There is a vast and ever‐accumulating amount of behavioural data on individually recognised animals, an incredible resource to shed light on the ecological and evolutionary drivers of variation in animal behaviour. Yet, the full potential of such data lies in comparative research across taxa with distinct life histories and ecologies. Substantial challenges impede systematic comparisons, one of which is the lack of persistent, accessible and standardised databases. Big‐team approaches to building standardised databases offer a solution to facilitating reliable cross‐species comparisons. By sharing both data and expertise among researchers, these approaches ensure that valuable data, which might otherwise go unused, become easier to discover, repurpose and synthesise. Additionally, such large‐scale collaborations promote a culture of sharing within the research community, incentivising researchers to contribute their data by ensuring their interests are considered through clear sharing guidelines. Active communication with the data contributors during the standardisation process also helps avoid misinterpretation of the data, ultimately improving the reliability of comparative databases. Here, we introduce MacaqueNet, a global collaboration of over 100 researchers (https://macaquenet.github.io/) aimed at unlocking the wealth of cross‐species data for research on macaque social behaviour. The MacaqueNet database encompasses data from 1981 to the present on 61 populations across 14 species and is the first publicly searchable and standardised database on affiliative and agonistic animal social behaviour. We describe the establishment of MacaqueNet, from the steps we took to start a large‐scale collective, to the creation of a cross‐species collaborative database and the implementation of data entry and retrieval protocols. We share MacaqueNet's component resources: an R package for data standardisation, website code, the relational database structure, a glossary and data sharing terms of use. With all these components openly accessible, MacaqueNet can act as a fully replicable template for future endeavours establishing large‐scale collaborative comparative databases.

Data collection biases are a persistent issue for studies of social networks. This issue has been particularly important in Animal Social Network Analysis (ASNA), where data are unevenly sampled and such biases may potentially lead to incorrect inferences about animal social behavior. Here, we address the issue by developing a Bayesian generative model, which not only estimates network structure, but also explicitly accounts for sampling and censoring biases. Using a set of simulation experiments designed to reflect various sampling and observational biases encountered in real-world scenarios, we systematically validate our model and evaluate it’s performance relative to other common ASNA methodologies. By accounting for differences in node-level censoring (i.e., the probability of missing an individual interaction.), our model permits the recovery of true latent social connections, even under a wide range of conditions where some key individuals are intermittently unobserved. Our model outperformed all other existing approaches and accurately captured network structure, as well as individual-level and dyad-level effects. Antithetically, permutation-based and simple linear regression aprroaches performed the worst across many conditions. These results highlight the advantages of generative network models for ASNA, as they offer greater flexibility, robustness, and adaptability to real-world data complexities. Our findings underscore the importance of generative models that jointly estimate network structure and adjust for measurement biases typical in empirical studies of animal social behaviour.

Animal social network analyses (ASNA) have led to a foundational shift in our understanding of animal sociality that transcends the disciplinary boundaries of genetics, spatial movements, epidemiology, information transmission, evolution, species assemblages and conservation. However, some analytical protocols (i.e., permutation tests) used in ASNA have recently been called into question due to the unacceptable rates of false negatives (type I error) and false positives (type II error) they generate in statistical hypothesis testing. Here, we show that these rates are related to the way in which observation heterogeneity is accounted for in association indices. To solve this issue, we propose a method termed the “global index” (GI) that consists of computing the average of individual associations indices per unit of time. In addition, we developed an “index of interactions” (II) that allows the use of the GI approach for directed behaviours. Our simulations show that GI: 1) returns more reasonable rates of false negatives and positives, with or without observational biases in the collected data, 2) can be applied to both directed and undirected behaviours, 3) can be applied to focal sampling, scan sampling or “gambit of the group” data collection protocols, and 4) can be applied to first- and second-order social network measures. Finally, we provide a method to control for non-social biological confounding factors using linear regression residuals. By providing a reliable approach for a wide range of scenarios, we propose a novel methodology in ASNA with the aim of better understanding social interactions from a mechanistic, ecological and evolutionary perspective.

Contraception is increasingly used to control wild animal populations. However, as reproductive condition influences social interactions in primates, the absence of new offspring could influence the females’ social integration. We studied two groups of wild macaques (Macaca fascicularis) including females recently sterilized in the Ubud Monkey Forest, Indonesia. We used social network analysis to examine female grooming and proximity networks and investigated the role of infant presence on social centrality and group connectivity, while controlling for the fertility status (sterilized N = 14, intact N = 34). We compared the ego networks of females experiencing different nursing conditions (young infant (YI) vs. old infant (OI) vs. non-nursing (NN) females). YI females were less central in the grooming network than other females while being more central in proximity networks, suggesting they could keep proximity within the group to protect their infant from hazards, while decreasing direct grooming interactions, involving potential risks such as kidnapping. The centrality of sterilized and intact females was similar, except for the proximity network where sterilized females had more partners and a better group connectivity. These results confirm the influence of nursing condition in female macaque social networks and did not show any negative short-term effects of sterilization on social integration.

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.

In social networks analysis, two different approaches have predominated in creating null models for hypothesis testing, namely pre‐network and node network permutation approaches. Although the pre‐network permutation approach appears more advantageous, its use has mainly been restricted to data on associations and sampling methods such as ‘group follows’. The pre‐network permutation approach has recently been adapted to data on interactions and the focal sampling method, but its performance in different scenarios has not been thoroughly explored. Here, we assessed the performance of the pre‐network and node network permutation approach in several simulated scenarios based on proneness to false positive or false negatives and with or without observation bias. Our results showed that the pre‐network permutation was sensitive to false positives in scenarios with or without observation bias. The node network permutation approach produced fewer false positives and negatives than the pre‐network approach, but only in scenarios without observation bias. In scenarios with observation bias, the node network permutation approach was outperformed by pre‐network permutation. Caution should be taken when using the pre‐network and node network permutations to create null models with data collected via focal sampling. This study provides future methodological research perspectives for social network analyses.