Stratified-sampling over social networks using mapreduce

Abstract

Sampling is being used in statistical surveys to select a subset of individuals from some population, to estimate properties of the population. In stratified sampling, the surveyed population is partitioned into homogeneous subgroups and individuals are selected within the subgroups, to reduce the sample size. In this paper we consider sampling of large-scale, distributed online social networks, and we show how to deal with cases where several surveys are conducted in parallel---in some surveys it may be desired to share individuals to reduce costs, while in other surveys, sharing should be minimized, e.g., to prevent survey fatigue. A multi-survey stratified sampling is the task of choosing the individuals for several surveys, in parallel, according to sharing constraints, without a bias. In this paper, we present a scalable distributed algorithm, designed for the MapReduce framework, for answering stratified-sampling queries over a population of a social network. We also present an algorithm to effectively answer multi-survey stratified sampling, and we show how to implement it using MapReduce. An experimental evaluation illustrates the efficiency of our algorithms and their effectiveness for multi-survey stratified sampling.