Vol.:(0123456789) 1 3
Social Network Analysis and Mining (2018) 8:35
https://doi.org/10.1007/s13278-018-0513-2
ORIGINAL ARTICLE
Effect of estimation method, definition of ratio, and the plausible
range in estimating social network size
Maryam Zamanian
1
· Farzaneh Zolala
2
· Ali Akbar Haghdoost
3
· Mohammad Reza Baneshi
4
Received: 6 January 2018 / Revised: 13 April 2018 / Accepted: 18 April 2018
© Springer-Verlag GmbH Austria, part of Springer Nature 2018
Abstract
A prime for network scale-up studies is the calculation of network size (C). To estimate C, respondents are asked about the
number of their acquaintances belonging to specific reference groups with known sizes. The aim of this manuscript is to
address influence of method of estimation and exclusion of unreliable reference groups on C. Recruiting 1275 women and
using 25 reference groups, C was calculated applying traditional and Means of Sums (MoS) estimators. This C is applied to
back-calculate the size of reference groups. To assess the closeness of back-calculated and real size, two types of ratio were
calculated: back-calculated over real size, and its reverse. The tolerable range for ratio was defined as (0.5, 1.5), (0.5, 2),
and < 1 based on the absolute logarithmic scale. The reference group corresponding to the poorest ratio was omitted. New
C was estimated based on the remainder of the reference groups. The whole process continued in an iterative fashion until
all ratios fall within the plausible range. In the traditional approach, C was robust with respect to definition of ratio and its
tolerable range. Minimum and maximum C were 174 and 186. In the MoS analysis, C values hugely diverse ranged 185–557.
This might partially be due to small number of eligible reference groups contributing in the estimation of C. As C is used to
estimate size of hidden groups, its calculation needs careful plan. We recommend authors to provide a range of values for C.
Keywords Network size · Estimator · Bias · Means of sums
1 Introduction
Estimating the size of hidden subpopulations is an important
issue in public health (Bernard et al. 2010). For example,
understanding and measuring the burden of HIV remains
facing several challenges around the globe. However, HIV
prevention, care, and treatment efforts including advocacy
for populations at risk of HIV, design and implementation
of national guidelines, and HIV program evaluations are not
feasible without relatively precise estimates of the impact
and magnitude of HIV among most-at-risk subpopulations
(for example, sex workers). Given the importance of inform-
ing public health policy makers with reliable information
about the number of people at risk of HIV, it is crucial to
quantify the population size of these key subpopulations.
Population size estimation (PSE) methods are divided
into direct and indirect methods. Direct PSE methods, such
as enumeration and survey, are prone to different biases
(Shokoohi et al. 2012). Using direct estimation technique
of population size has various challenges such as hallmark
and discrimination. In addition, survey for direct population
size estimation (according to the relatively small population
size) requires a large sample size, and it is not often feasible.
Network scale-up (NSU) is an indirect PSE method which
is implemented in general population. While in direct survey,
respondents are asked to provide information about them-
selves, in NSU method respondents are asked to concentrate
on their network (shown by C) and to provide the number
of their acquaintances who belong to the subpopulation
* Mohammad Reza Baneshi
rbaneshi2@gmail.com
1
Department of Epidemiology, School of Health, Arak
University of Medical Sciences, Arak, Iran
2
Social Determinants of Health Research Center, Institute
for Futures Studies in Health, Kerman University of Medical
Sciences, Kerman, Iran
3
HIV/STI Surveillance Research Center, and WHO
Collaborating Center for HIV Surveillance, Institute
for Futures Studies in Health, Kerman University of Medical
Sciences, Kerman, Iran
4
Modeling in Health Research Center, Institute for Futures
Studies in Health, Kerman University of Medical Sciences,
Kerman, Iran