The topic discusses the fundamental principles of validity and reliability as applied to computational social network data. Measurement error and sampling bias are analyzed as primary sources of distortion in network representation, distinguishing between omissions (missing edges or nodes) and commissions (false inclusions). Class exercises demonstrate how perturbations in the observed graph alter key metrics such as density and average shortest path, revealing the fragility of inference in incomplete datasets. Reliability is further examined through temporal analysis of dynamic networks, where measures such as the Jaccard Index and Quadratic Assignment Procedure (QAP) are used to test structural stability across time slices. 

Distinguishing genuine social evolution from noise is achieved through permutation testing and null distribution analysis. The concept of triangulation is introduced to emphasize methodological redundancy as a safeguard against misinterpretation, and methodological pluralism is presented as a strategy for robust structural inference. Together, these discussions establish that the credibility of network analysis rests not only on computation but also on critical evaluation of its data and methods.

Learning Outcomes

Guide Questions

Topic Outline

Reading List