Network data in the Partial Least Squares framework

This thesis stems from the idea to draw a statistical soft-modeling framework to network data. Network data arise in very different and multidisciplinary fields in order to study relational ties among units. The different fields highlighted in recent years the necessity to collect relational and attribute data, as well as metadata describing the actors in the network. Since usual relational datasets are characterized by i) very different amount of units (from very few units to huge networks), ii) biased sampling (for instance, people with more social connections may have a higher chance of selection) and iii) a kind of heterogeneous information attached to both nodes and ties; these facets highlight the peculiarity for classical statistical tools and models to be applied. In the specific, we are interested in processes where social relations provide a basis for the alteration of an attitude or behavior by one actor in response to another one. This social process of attitude change, that appears in a social network, is known as social in uence or contagion. A mathematical formalization of the effects of social network on behaviors is given by the Network Effects Model. From an empirical point of view, these models are far from being directly observable. The possibility of measuring them as latent factors depending from multidimensional constructs still remains. All together, a component-based approach to network data through Partial Least Squares-path model algorithms is proposed. A simulation study is presented.