So now we're going to talk about more specifics of social networks analysis. And so, first of all, social network analysis can be broken down into two different perspectives. One is a personal network perspective and the second is a complete network perspective. So the personal network perspective, or egocentric networks, is focused on the social networks that surround a focal person. So this is very much from the individual's perspective. So in this instance you're focused on an individual, or ego, and the network members that surround them. And so the data collection for this type of a perspective often involves asking a person about who they're socially connected to, and the characteristics of these network members, and the types of relationships that they have with them. And so you see a network where the ego at the center surrounded by their network members that look like they're spokes coming off of a tire. And complete network approaches, or they're also called sociocentric networks, is an approach that's much more focused on a community or group level. So in this type of analysis, we focus on defining a particular social group and defining a boundary around that social group. And that boundary might depend on our goals or our research questions. And then in this approach we measure the relationships among all of the members of that social group, so it's really a consensus of the relationships among the group members. And so the data collection for this strategy involves defining your social group, and then going in and getting a measure of the relationships among all group members, and looking at a map of this social network among the community as a whole. So this approach is more suited to something like looking at all of the relationships among students in a school or the relationships among people in an organization or all of the members of a particular village. Or like the graphic on the bottom right, it might be all of the connections on Facebook of people all around the world. So social network analysis is very focused on social network structure, and mapping nodes and ties. So nodes represent social actors. These are the dots that we see in these network maps. And lines, or ties between nodes, represent the relationships between these people, and then these relationships might be directed, or they might be undirected. So in a directed relationship, this might be something where one node nominates the other node as a friend, but that might not be reciprocated. Where a non-directed relationship might be something like a sexual partnership, where we don't need to specify the direction. And relationships can have many qualities. They might be strong or weak, and they might represent a whole bunch of different types of relationships. So things like friendship, or acquaintance, or even aggressive relationships. So when we start doing social network analysis, we need to go and get social network data. And typically, we want to gather data and information about the actors in our network and also information about the relationships among the actors. And then this can be collected in a number of different ways, possibly through observation or through data mining, things like gathering data off social media networks or gathering phone records. And it's also commonly collected through a process that uses name generators, and this is where we ask people to tell us about who their social connections are. And then we ask them a number of questions about the characteristics of those network members and the types of relationships that they have with them. And then we can take this relational data and we can create matrices of these relationships, and in these matrices each cell represents the presence or the absence of a relationship between each pair of people in the network. And then social network analysis provides us with a huge set of tools that allows us to visualize these networks and also to summarize the structure of the networks and the locations of particular nodes in these networks. And just like traditional statistics, networks statistics give us a range of both the descriptive and inferential statistics. So descriptive statistics can be used to summarize the characteristics of a network or a group of networks that you've observed. For example, you might want to summarize the average number of friendship nominations that nodes tend to receive or the distribution of these number of friendship nominations. Inferential statistics let us test network hypotheses. So these might be questions like, is it more likely that we see a tie among nodes for the same gender, or who have the same types of characteristics? Or are the attributes or behaviors of nodes impacted by the people that they're connected to in the network? So networks at their most basic level are made up of dyads, and so this is the relationship between pairs of people. And there's a number of characteristics of dyads that we're often interested in measuring and assessing in social network analysis, and one of these is reciprocity. And this is the characteristic of a dyad that captures whether or not the relationship is mutual or more one-directional. So an example of this is whether the friendship nominations between a pair of people is just from one person, and whether or not it's reciprocated by the other. And these can reflect different types of hierarchies and different kinds of exchanges within the network. A second characteristic is the the intensity of the relationship and this could be thought of in a number of different ways. It might simply reflects the frequency of interaction between a pair of people, or an emotional intensity in the relationship, where they might characterize a relationship as being very strong or weaker and more of an acquaintance. And lastly, we want to think of these relationships as being potentially very complex and that there maybe is a number of different roles and aspects to the relationship between a dyad. So we could think of a pair of adults, who may say that they're friends with each other. And in one pair of adults, they really don't share any other type of role or relationship other than friendship. But we may observe another pair of adults who are friends with each other, but also share some additional roles and complexities in their relationship. One of the pair is the other person's boss, and so when we delve deeper into the complexities of their relationship, we find that there is some uni-directional or directed relationships of trust and advice giving that make this relationship different to the one we observed that was just a friendship. So when when we aggregate all of this dyadic information into a larger social network, social network analysis provides us with the set of tools to understand some of the emergent characteristics of these networks. A number of them are listed here, and these are characteristics that are often of interest to researchers in terms of understanding social behavior and social phenomenon. One is simply network size. This is just the number of nodes or people that are in a social network. So we may think of a social network as the hundred people that we know really well, or we might be looking at a social network of a thousand students in school and trying to understand relationships and complex relationships among that larger network. Another characteristic is network density, and this is the number of relationships that are in the network relative to the total number of possible ties, meaning that if all of the people and all of the pairs in the network were connected to each other. So one way to think of this is to think about the start of a class in the start of the semester, and coming in and not knowing very many people in your class. And the density in that network at the start of the semester may be quite low. There may be just a few pairs of people who are friends with each other, relative to the total number of possible pairs in that classroom. But typically what we see over the course of a semester is the density of these friendship networks increases, so that by 6 months after you've been spending time in class together, more friendships have formed and we see more pairs of friendship relationships among the network and a greater network density as a result. And density is often a really interesting phenomenon because we know it helps to explain the flow and diffusion of information and ideas and even disease through a network. Another characteristic is network clustering, and this is the phenomena where densely connected subgroups tend to form within a network. I mean, we often think of this in terms of small cliques or clusters of people who are more closely connected to each other than they are to the rest of the network. Network composition is a characteristic where we can describe the types of people that make up the members of a network. So we might think about the describing the proportion of the network that's female. Network homogeneity is a characteristic that describes the similarity among people in a network, in particular, the similarity among people who are socially connected. And so in networks we often observe that social ties and connections tend to form among people who share similar demographics, similar gender or age or socioeconomic status, and similar race and ethnicity. And lastly, geographic dispersion is a characteristic that helps us to think about how close or far away from each other in terms of geographic space people are in a social network and how that might impact on their social connections. Social network analysis also, beyond thinking about just the structure of the network and describing it, really helps us to think about the position and roles that individuals play in these social networks, and thinking about that in terms of their location in this social structure. And so one thing we often focus on is people's membership in groups or clique within a network. So these are these local, more densely connected sub groups in a network that we tend to see. And these are often characterized by this homophily feature, where we see this trend for birds of a feather to flock together. And these densely connected social groups to be people who share a lot of similar qualities or behaviors or characteristics. Within these small groups are social cliques. We also see evidence of social influence where people are very influenced by these densely connected subgroups that they're in. We also want to think about central members or opinion leaders within the network. Often we think about this in terms of nodes in the network who have the most social connections. And these are often the people who we would think of as being the most popular, potentially the most influential in a network. And for health interventions, we really think about these people as important agents for change. And so, in the network on the right, which is a network of friendships among a group of students, we're able to see these sort of clustering and small grouping of people. And we might think about the role that these people play in social networks. I've gotta do this one again, sorry. I stuffed up the animations. Should I just start? >> Yeah go ahead, whenever you're ready. >> Okay. So social network analysis, let me, I'll put a break there, social network analysis also gives us a set of tools to think about the position of individual nodes in a network and their location in these social network structures. So one thing we often want to think about is people's membership in these groups or cliques that I talked about on the last slide. So these are these local, more densely connected subgroups that we often observe in networks. And you can see some examples of that on the graph on the right. And often these subgroups are characterized by homophily. So this is the tendency for birds of a feather to flock together. And this tendency for people who are similar in a range of demographic characteristics like gender or age or socio-economic status, or even race and ethnicity, to be more closely socially connected to one another in networks. It is often in these social groups where we see a lot of social influence, where the behaviors and the norms and ideas of the people that are in these densely connected social groups really influence one another. We also often want to think about central members in the network, what we might think of as opinion leaders. These are members who tend to have the most social connections. And we often think about them as people who are popular and potentially influential, and in health interventions we think of these people as important agents for change. And there's an important interdependence between these opinion leaders and these social group clusters. Often these central members or these opinion leaders are embedded in these smaller social groups. And again, we want to think of this relationship as quite bidirectional and quite a system that evolves over time, where central members can be quite influential in these social groups, and sort of dictate the norms and behaviors that are sort of socially desirable in those groups. But they're also very much influenced by the groups that they're in and they have a pressure to maintain their social status, by not deviating from what the group thinks is normal or cool or okay. So on this slide I want to give you an example of a real life adolescent friendship network. This is a group of students that we observed who were in grade eight. And in this study we asked students in this school to tell us of among their classmates who their best friends were. And so we're able to map this network that you see on the screen. We also asked them about how much physical activity they've done in the past week. And we were interested in how the network might be influencing the physical activity of these youths. So when we look at this network and map it, we're able to see quite segmented clusters or more densely connected clusters and pockets of friends within this friendship network. And we start to see some tendencies of homophily in the network. So this tendency for people who are socially connected to be more similar than we might expect by chance in the behavior that we're focused on. So here you can see pockets of students who are quite physically active, and then other pockets, although there is some mixing. But other pockets of students where you see more red and yellow nodes, who are some of the least active students. And so an intervention that we might think about in this network might focus on identifying opinion leaders within these social clusters, and thinking about how we can target these opinion leaders to get them to be more engaged in physical activity, to encourage their close friends to be physically active. And think about diffusing our intervention through the network utilizing the power of some of these opinion leaders.
