Validity Issues in the Use of Social Network Analysis with Digital Trace Data

Abstract

Abstract

The empiricist’s challenge: Asking meaningful questions in political science in the age of big data

“…This deluge comes in the form of digital trace data-data wittingly or unwittingly produced in the context of using digital services or devices (Howison, Wiggins, & Crowston, 2011). These data hold great promise for business, government, and academia as they potentially document individuals' behavior unfiltered by obtrusive or unreliable measurements-such as self-reported behavior or observations in artificial laboratory environments.…”

“…By offering a precise and unobtrusive documentation of user behavior, digital trace data potentially allow social scientists to avoid three traps of other measurement approaches (Golder & Macy, 2014;Howison et al, 2011;Salganik, 2017). For researchers interested in respondents' behavior and attitudes, reliance on self-reported behavior-as in surveys-is notoriously problematic.…”

How healthy are software engineering conferences?

“…Identity merging across multiple data sources has been explicitly recognized as a reliability threat [66]. Automatic merging techniques might have missed multiple representations of the same individual or considered multiple individuals to be one and the same person.…”

Validating ScienceU+0027s Power Players: Scientometric Mixed Methods for Data Verification in Identifying Influential Scientists in a Genetics Collaboration Community

“…3 Verifying CNA methods: Hubs Trace data has a track record of falling prey to validity worries (Howison, 2011). Drawbacks of complex social network analysis include, but are not limited to, confirming interpretation of findings (e.g., understanding sequencing submission processes and community detection vagaries) and the retrospective and temporal nature of the data.…”

“…Known drawbacks of complex network analysis (CNA) have been discussed by (Howison, 2011), while methodological and technical challenges of database design and ensuring data quality were described by (Qin, Costa, & Wang, 2015). They underscore the importance of mixed methods, especially given the emergence of new forms of intellectual property units, knowledge products yet to be rigorously defined.…”