Global Network Inference from Ego Network Samples: Testing a Simulation Approach

Abstract

Network sampling poses a radical idea: that it is possible to measure global network structure without the full population coverage assumed in most network studies. Network sampling is only useful, however, if a researcher can produce accurate global network estimates. This article explores the practicality of making network inference, focusing on the approach introduced in Smith (2012). The method uses sampled ego network data and simulation techniques to make inference about the global features of the true, unknown network. The validity check here includes more difficult scenarios than previous tests, including those that go beyond the initial scope conditions of the method. I examine networks with a skewed degree distribution and surveys that limit the number of social ties a respondent can list. For each network/survey combination, I take a random ego network sample, run the simulation method, and compare the results to the true values (using measures of connectivity and cohesion). I also test the method on local measures of network structure. The results, on the whole, are encouraging. The method produces good estimates even in cases where the degree distribution is skewed and the survey is strongly restricted. I also find that is it better to not truncate the survey if possible. If the survey must be restricted, the researcher would do well to infer the missing data, rather than use the raw data naively.

Using Ego Network Data to Inform Agent-based Models of Diffusion

“…Such "forgetting" will distort the number of alters listed and thus the inferred degree distribution. Similarly, many studies truncate the number of alters one can list, but this too can lead to a distortion in the degree distribution (Smith 2015).…”

“…We describe the approach in general terms here, but see Smith's (2012Smith's ( , 2015 study for technical details.…”

“…Inference is harder when the degree distribution is skewed: a few actors have disproportionately high degree, yet they are no more likely to be sampled than any other node. High degree nodes thus have a large impact on network structure, but are often missed in a random sample (see Smith 2015). And third, the network is disconnected, with 20% of the nodes outside the main component (a component is a set of nodes connected by at least one path; the main component is the largest set of nodes connected by at least one path).…”

“…Past work has tested a sampling approach on the features of the network (e.g., can we use ego network data to estimate transitivity?) but has not considered diffusion processes directly (Smith 2015). Such tests miss the behavioral, stochastic component at the heart of an ABM, where different adoption rules (i.e., under what conditions will an actor adopt a new product?)…”

Modeling Network Dynamics

“…We discuss the capacity for each approach to handle other types of relations. We also assume "whole" network data, where relations between all nodes within a specified context are measured, though studies with other types of samples, such as ego-centered network data, have adopted the methods we review here (Krivitsky & Morris 2015;Marcum & Butts 2015;Smith 2012Smith , 2015Lubbers et al 2010). As with any method, whether these approaches are the ideal choice depends on one's research question and data.…”