Respondent Driven Sampling for Co-location Networks

Dr. Catherine Calder, Department of Statistics
Rank at time of award: Associate Professor
and
Dr. Christopher Browning, Department of Sociology
Rank at time of award: Professor

 

 

This proposed research project seeks to develop and pilot an alternative sampling strategy for co-location networks. Instead of the traditional approach of sampling individuals from the population and learning about only the locations they report, we propose a two-step design in which a random sample of individuals is selected and these subjects recruit additional individuals present at their activity locations into the study. This general type of design is known as respondent driven sampling (Heckathorn 1997), and it has been successfully implemented for one-mode network sampling of hard-to-reach populations (Heckathorn et al. 2002; Iguchi et al. 2009; Rudolph et al. 2011). Instead of the hard-to-reach population setting, we propose to use respondent driven sampling for an entirely different purpose – to better learn the local spatial structure of a co-location network. Consistent with very recent work in the statistics literature (Shalizi and Rinaldo 2013; Schweinberger and Handcock 2014) and based on our preliminary findings, it is apparent that accurately estimating such local structures in networks is essential for making valid inferences. The proposed respondent

driven sampling approach captures the local structure of co-location networks more accurately and cost effectively than the traditional one-mode sampling design.