Specialized Session 9: Social Data Science: Relational and Structured Data

In this paper, we propose a framework for providing and evaluating predictions under cluster structure assumption and missing data. In particular, we assume that testing set units have missing values for some covariates and define the donors for the units in the test set according to a similarity definition based on the predictive model. For this aim, we employ clusterwise regression and evaluate the predictive performances under several existing missing data imputation approaches. We apply the proposed framework to predict the amount of the received green founds of Italian SMEs. Our results show the usefulness of the cluster-wise regression for both donors’ identification and prediction tasks.

This study explores the use of DISTATIS, a three-way multidimensional scaling technique, for analysing attributed networks. By integrating both network topology and node attributes (including qualitative and quantitative features), DISTATIS provides a comprehensive framework for studying the coherence between structure and attributes. A simulation study was conducted on stochastic block model networks, progressively disrupting their structure through edge rewiring while preserving attribute distributions. Results demonstrate that DISTATIS effectively captures variations in coherence, highlighting its potential for analysing social and complex networks and understanding the interplay between structural and attribute-based information.

This paper discusses practical issues related to data collection and sampling strategies for evaluating secondary data sources using web scraping techniques and automated classification methods. It highlights the significance of assessing the reliability of data providers and the methodologies used in their data collection, which directly impacts the eligibility of sampling units in e-commerce surveys, making relevant implications on the cost and the quality of any survey conducted by National Institutes of Statistics.