The research topic of this study regards scientific collaborations between university researchers. In the sociology of science approach three research streams are prevalent. Micro perspective focus on interactions in laboratories in order to uncover the impact of social and cultural norms in the everyday production of scientific results. Other studies investigates on different factors may cause scientists to entertain collaborative relationships. Macro perspective analysis the influence of scientific organizations on society and the inverse. Meso-perspective holds together two levels of analysis and focus interactions within and between scientists and the institutions they work in. Other studies, employing spatial interaction models, attempts to describe and analyze the collaboration network to understand or how the geographical distance between actors influence collaboration performance or to detection of the community structure. Objective: The objective of this research is part of a larger project aimed to find the factors that explain changes of scientific communities over time. In this paper we focus our attention at meso level on scientific collaborations between Italian university researchers to detect communities. Data: The data used in this study cover 10 years (1999-2008) of Scientific Project of National Interest (PRIN) from the Italian Ministry of Instruction, University and Research. Methodology: We employ the techniques of the Social Network Analysis and, in particular, we use original community detection algorithm. Dynamic communities can be defined as a group of nodes which stay connected over time. Various algorithms in community detection was proposed in literature. Each different method could be applied to different network type, with different data structures in order to identify the communities. However, results can differ from an algorithm to another. It could be necessary an approach which allows to detect the relevant communities as each method suggest. To this aim we use an algorithm based on three steps: first, we detect some communities stable over time which show relevant characteristics of robustness; secondly, we look for patterns of cooperation over time (by considering the co-participation matrices); finally, we take into account the structural changes occur in the network evolution. The proposal brings together different methods in order to stabilize the results. In this case the different methods ensure the capacity to detect the different cooperation patterns which can occur in the networks. By considering the nature of the data where longitudinal we can to obtain the co-participation matrices which measure the level of cooperation of the single nodes over time. It is important to note that this participation could be related different temporal interval. So the co-participation matrices can measure the cooperation over different intervals in time. Results: The communities show relevant cooperation patterns between the universities and the research groups. In particular it is possible to observe a size effect between the research groups: the university size, relevance and also prestige seems to affect the mechanisms of the creation or destruction of the communities.
- Social Network Analysis,
- Temporal Network Analysis,
- Community Detection,
- PRIN Networks,
Available at: http://works.bepress.com/carlo_drago/104/