Investigating the Variation of Personal Network Size Under Unknown Error Conditions

Abstract

This article estimates the variation in personal network size, using respondent data containing two systematic sources of error. The data are the proportion of respondents who, on average, claim to know zero, one, and two people in various subpopulations, such as “people who are widows under the age of 65” or “people who are diabetics.” The two kinds of error—transmission error (respondents are unaware that someone in their network is in a subpopulation) and barrier error (something causes a respondent to know more or less than would be expected, in a subpopulation)—are hard to quantify. The authors show how to estimate the shape of the probability density function (pdf) of the number of people known to a random individual by assuming that respondents give what they assume to be accurate responses based on incorrect knowledge. It is then possible to estimate the relative effective sizes of subpopulations and produce an internally consistent theory. These effective sizes permit an evaluation of the shape of the pdf, which, remarkably, agrees with earlier estimates.