|
Activity Number:
|
390
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Social Statistics Section
|
| Abstract - #309544 |
|
Title:
|
Estimating Personal Network Size
|
|
Author(s):
|
Tyler H. McCormick*+ and Tian Zheng and Matthew J. Salganik
|
|
Companies:
|
University of Connecticut and Columbia University and Columbia University
|
|
Address:
|
207B Wu, Storrs, CT, 06269,
|
|
Keywords:
|
Social networks ; Network size ; Sampling ; Overdispersion
|
|
Abstract:
|
Social network size provides information about the structure of relationships between individuals. This article presents a means of estimating network size and demonstrates how to design studies that estimate network size efficiently. We propose an estimation procedure based on count data reflecting the number of individuals with a particular first name known by the respondent. Estimation is done using a multilevel overdispersed Poisson regression model that reduces bias by accounting for the relationship between a respondent's age and the popularity of names over time. Mixing coefficients, which represent the proportion of a respondent's network that is of a particular age, are also estimated. We then derive simple estimates which, with appropriate age profiles, are self-weighting. Suggestions are given for survey designs that facilitate efficient estimation of network size.
|
