The Fit Between Theory and Data in Respondent-Driven Sampling: Response to Heimer

Abstract

., thisissue). As Heimer states, Respondent-Driven Sam-pling (RDS) is “an innovative and powerful method-ology,” thus, we welcome this opportunity to furtherclarify the theoretical basis and applications of RDS.Like any other probability sampling method,RDS is based on a statistical theory of the samplingprocess. Such theory provides the means to calcu-late population estimators of minimal or zero bias,and estimates of the variability of those estimatorsin the form of confidence intervals or standard er-rors. This theory also constrains the contexts in whichsampling can validly take place, because necessaryinformation required by the statistical theory must beattainable, and data structures must have the formpresumed by the statistical theory. Consequently,evaluating an application of a sampling method mustassess the fit between the requirements of the sam-pling method’s statistical theory and the data fromwhich research conclusions are derived. To the ex-tent that the assumptions of the statistical theory areviolated, confidence on the validity and reliability ofestimators is correspondingly reduced. This is the fo-cus of Heimer’s comments. He repeatedly expressesconcerns that we failed to address what he sees asdiscrepancies between the data reported in our arti-cle and the requirements of the RDS method. These