What Big Data tells: Sampling the social network by communication channels

Area of Science:

• Social Network Analysis
• Big Data Analytics
• Network Science

Background:

• Big Data is crucial for understanding large-scale societal structures and dynamics.
• Social interactions form a multiplex network, with each layer representing a communication channel.
• Typically, data is available for only a subset or sample of these channels.

Purpose of the Study:

• To introduce a model explaining how communication channel selection affects social network properties.
• To investigate the impact of sampling on network characteristics like degree distribution and assortativity.
• To analyze the broader implications of these sampling-induced changes.

Main Methods:

• Developed a model based on a natural bilateral communication channel selection mechanism.
• Analyzed changes in network properties resulting from single-channel data sampling.
• Investigated effects on degree distribution and network assortativity.

Main Results:

• The model demonstrates how sampling leads to a monotonically decreasing degree distribution, contrary to whole-network expectations.
• Observed that assortativity can emerge or be amplified due to the sampling method.
• Quantified consistent changes in network properties attributable to the selection mechanism.

Conclusions:

• Sampling methods significantly alter the observable properties of social networks.
• The proposed model provides a framework for understanding discrepancies between single-channel and multiplex network analyses.
• Findings have implications for interpreting Big Data from social interactions and network science research.