SOCI 2001 week 2 Discussion Essay
Social Network Analysis
Social Network analysis is the study of structure, and how it influences health, and it is based on theoretical constructs of sociology and mathematical foundations of graph theory. Structure refers to the regularities in the patterning of relationships among individuals, groups and/or organizations. When social network analysis is undertaken, the underlying assumption is that network structure, and the properties of that structure have significant implications on the outcome of interest.SOCI 2001 week 2 Discussion Essay
Due to its focus on network structure rather than individual characteristics and or behaviors of network members, the data required for appropriate analysis differs from what is typically collected in non-relational epidemiological study designs. Typically, study designs that focus on individual characteristics/behaviors and how those characteristics influence health, collect and conduct analysis on attribute data. Attribute data is defined as data that reflects the attitudes, opinions, and behaviors of individuals or groups. Conversely, social network analysis requires not only attribute data, but is built on the collection and analysis of relational data. Relational data refers to contacts, ties and connections, which relate one agent in a network to another. Relational data cannot be reduced to properties of the individual agents themselves but to a system/collection of agents.SOCI 2001 week 2 Discussion Essay
Description
The majority of social network studies use either whole (Socio-centric) networks or egocentric study designs. Whole network studies assess relationships between individuals or actors that for analytical purposes are regarded as bounded or closed, even though in actuality the boundaries of the network are in fact permeable and/or ambiguous. When whole network studies are conducted, the focus of the study is to measure the structural patterns of how individuals within the network interact and how those patterns explain specific health outcomes. The underlying assumption made when whole network analysis is conducted, is that individuals that make up a group or social network will interact more than would a randomly selected group of similar size.
In a sociology-centric study, members of the network are usually known or are easily determined because the focus is usually on closed networks that are a priori defined. For this reason, data collection for sociology-centric network analysis involves enumerating all network members, and administering saturation surveys to all network members. A saturation survey provides respondents with a roster of all network members, and respondents are asked to identify members with whom they are affiliated. From this data, actor-by-actor matrices can be constructed and social network analysis can be conducted.
When the network of interest does not have clearly defined boundaries, sociology-centric studies result in snowball or respondent driven sampling to generate the network and collect data to identify structural patterns. In respondent driven sampling, a small number of network members are interviewed and asked to name other network members, and those named members are also interviewed and asked to name other network members. This iterative process is continued until all network members are identified, or for an a priori set number of waves established before study initiation. The assumption made when respondent driven sampling is used is that the sampled network is representative of all other segments of the network from which data has not been collected. Respondent driven sampling uses name generator surveys to identify network members, followed by name interpreter questions to solicit information about the named actors, their characteristics, and relations to the focal actors.SOCI 2001 week 2 Discussion Essay
Egocentric network designs, on the other hand, focus on a focal actor, ego, and the relationships between the ego and named actors or objects within their social networks. These types of designs collect data on the relationships involving the ego and the objects, alters, to which they are linked. Egocentric study designs use either name generators or position generators to obtain both attribute and relational data that can be used to construct actor-by-actor from which egocentric data analysis can be constructed. Position generators are used to identify people who fill particular value rolls, such as lawyers, where as name generators, as discussed above, are questionnaires that ask the ego questions about individuals to whom he or she is connected in a specific way. Unlike in sociology-centric studies, however, resource constraints preclude the subsequent interview of named alters, and therefore the ego serves as the informant for not only their own relationships with the alters, but also the alters relationships with each other. Name generator questions like in socio-centric respondent driven sampling are usually followed by name interpreter questionnaires.SOCI 2001 week 2 Discussion Essay
Analysis of Social Network Data
Network data, though collected at the level of the individual, is analyzed at the structural level. Data is organized as an actor-by-actor matrix as depicted in figure 1B. Data as displayed in figure one depicts the presence or absence of a tie. When the strength of a tie is also of interest, i.e. valued data, similarity or distance matrices could be used. Similarity matrices depict stronger ties with increasing numerical values, while increasing numerical values in distance matrices reflect weakened ties because the greater the distance between two actors, the weaker the ties. Any actor-by actor matrix can be converted into graphs and analyzed using social network analysis software such as UCINET.
Graphs are visual representations of a network. Actors within a network are displayed as nodes and the lines connecting nodes are representative of the ties between two actors. Graphs can be directed, indicating the relationship is directed from one agent to the other, or valued, indicating the strength of the tie. Though, visualizing the data is informative, the crux of social network analysis lies in the calculation of descriptive measures that reveal important characteristics about 1) position of network actors, 2) properties of network subgroups, and 3) characteristics of complete networks. SOCI 2001 week 2 Discussion Essay