Inferring Structural Characteristics of Networks with Strong and Weak Ties from Fixed-Choice Surveys

Naghmeh Momeni, Michael G. Rabbat

Abstract

Knowing the structure of an offline social network facilitates a variety of analyses, including studying the rate at which infectious diseases may spread and identifying a subset of actors to immunize in order to reduce, as much as possible, the rate of spread. Offline social network topologies are typically estimated by surveying actors and asking them to list their neighbours. While identifying close friends and family (i.e., strong ties) can typically be done reliably, listing all of one's acquaintances (i.e., weak ties) is subject to error due to respondent fatigue. This issue is commonly circumvented through the use of so-called "fixed choice" surveys where respondents are asked to name a fixed, small number of their weak ties (e.g., two or ten). Of course, the resulting crude observed network will omit many ties, and using this crude network to infer properties of the network, such as its degree distribution or clustering coefficient, will lead to biased estimates. This paper develops estimators, based on the method of moments, for a number of network characteristics including those related to the first and second moments of the degree distribution as well as the network size, using fixed-choice survey data. Experiments with simulated data illustrate that the proposed estimators perform well across a variety of network topologies and measurement scenarios, and the resulting estimates are significantly more accurate than those obtained directly using the crude observed network, which are commonly used in the literature. We also describe a variation of the Jackknife procedure that can be used to obtain an estimate of the estimator variance.

Introduction

Mapping offline social networks is a notoriously difficult task. Unlike digital platforms where interactions are logged automatically, researchers studying physical communities must rely on surveys. However, human memory and patience are limited. To mitigate "respondent fatigue," researchers often employ fixed-choice surveys, capping the number of acquaintances a participant can list (e.g., "name up to 5 friends").

While practical for data collection, this truncation creates a "crude" observed network that is significantly sparser and less clustered than the ground truth. Using such biased data to model phenomena like viral spread or information diffusion leads to inaccurate predictions. This research presents a statistical framework to "de-bias" this data and infer the true structural characteristics of the underlying network.

Methodology: The Two-Layer Model

The core of the proposed approach is a generative model that distinguishes between tie types. The challenge is degree truncation: while respondents (seeds) generally report strong ties (close friends and family) accurately, they often fail to list all weak ties (acquaintances) due to the fixed-choice cap.

The two-layer network model illustrating strong ties (top) and weak ties (bottom) among seeds and non-seeds.

The model treats the social network as two layers. A set of "seeds" is chosen via a Bernoulli process. These seeds report all their strong ties but only a fixed number ($B$) of their weak ties, selected uniformly at random. By using the method of moments, the authors derive estimators for the total population size ($N$), the sampling probability ($q$), and the average degrees ($K_s, K_w$) based on the observed links between seeds and non-seeds.

Clustering and Higher-Order Moments

Estimating the clustering coefficient is particularly complex because truncation hides many triangles. The authors analyze how triangles and "triads" (three-node motifs) in the real network are likely to appear in the truncated survey data.

Examples of open triads with different combinations of strong (solid) and weak (dashed) ties.

By identifying the various ways a triangle or open triad can be partially observed (e.g., two strong links, or one strong and one weak link as shown in the figure), they can back-calculate the true counts of these structures. This allows for a much more accurate estimation of the network's density and social cohesion than calculating metrics directly from the raw, "crude" survey data.

Experimental Validation

The estimators were validated using both synthetic models (Watts-Strogatz, Barabási-Albert, Holme-Kim) and real-world data from the "Villages" dataset in rural India.

Key findings include:

  • Accuracy: Estimators for $N$, $q$, and average degrees were remarkably close to true values, with bias decreasing as sample size increased.
  • Superiority: The method significantly outperformed standard techniques that do not differentiate between tie types.
  • Efficiency: The entire inference procedure for a network of approximately 10,000 nodes takes about one second on a standard laptop.

Conclusion

This work demonstrates that accounting for tie heterogeneity is essential for accurate network inference. By acknowledging that surveys treat strong and weak ties differently, researchers can recover the "full picture" of a community's structure from limited data. Future work may relax the assumption of uniform weak tie selection and extend these methods to individual-level metrics like betweenness centrality.

Summary updated

Abstract

Knowing the structure of an offline social network facilitates a variety of analyses, including studying the rate at which infectious diseases may spread and identifying a subset of actors to immunize in order to reduce, as much as possible, the rate of spread. Offline social network topologies are typically estimated by surveying actors and asking them to list their neighbours. While identifying close friends and family (i.e., strong ties) can typically be done reliably, listing all of one's acquaintances (i.e., weak ties) is subject to error due to respondent fatigue. This issue is commonly circumvented through the use of so-called "fixed choice" surveys where respondents are asked to name a fixed, small number of their weak ties (e.g., two or ten). Of course, the resulting crude observed network will omit many ties, and using this crude network to infer properties of the network, such as its degree distribution or clustering coefficient, will lead to biased estimates. This paper develops estimators, based on the method of moments, for a number of network characteristics including those related to the first and second moments of the degree distribution as well as the network size, using fixed-choice survey data. Experiments with simulated data illustrate that the proposed estimators perform well across a variety of network topologies and measurement scenarios, and the resulting estimates are significantly more accurate than those obtained directly using the crude observed network, which are commonly used in the literature. We also describe a variation of the Jackknife procedure that can be used to obtain an estimate of the estimator variance.
BibTeX
@article{momeni2017inferring,
  author = {Naghmeh Momeni and Michael G. Rabbat},
  title = {{Inferring Structural Characteristics of Networks with Strong and Weak Ties from Fixed-Choice Surveys}},
  journal = {IEEE Trans. Signal Inf. Process. Over Networks},
  year = {2017},
  url = {https://openreview.net/forum?id=-3z4tI7fhm6},
}
Figure