Web Survey Bibliography

Title Sub-sampling of Alters as a Mean to Reduce the Respondent Burden in Egocentric Social Network Questionnaires
Author Kronegger, L., Vehovar, V., Hlebec, V., Koren, G.
Year 2004
Access date 28.08.2004
Abstract In egocentric social network research each responding person (i.e. ego) typically reports the details for all persons (i.e. alters) from his/her network. This poses a considerable burden to the respondents. In interview surveys (telephone, face-to-face) this also increases the surveys costs. that order such survey. With Web surveys and self-interviewing the burden is almost entirely shifted to the respondent, who can easily quit the survey process at any stage. As a consequence, an increased response burden can introduce severe problems to survey data quality (nonresponse, measurement errors). Some studies have already shown that dropout rates inWeb surveys significantly increase with the increased size of the personal network. The aim of the paper is to examine a simplification of tipicaly the measurement tool for attaining collecting data on egocentric social networks. Instead of, say, 15 alters that certain person actually reported in name-generation process, we ask the detailed questions for only 5 alters. We thus study the pay-offs between the corresponding loss of information on one side and the benefits arising from reduced respondent burden on the other side. First, the sub-sampling simulations were performed based on real parameters. The framework of two-stage cluster sampling was applied. Different (sub)sampling approaches showed considerable discrepancy in nonresponse bias and also in precision. The use of rank set sampling seemed to be an especially promising strategy. In particular, the number of sub-sampled alters and the type of the sampling mechanism were the most important parameters. In the empirical part the procedure was tested with egocentric networks data from face-to-face, telephone and Web surveys in Slovenia 2001-2003. The results confirmed that in presence of the attrition this sub-sampling approach can be costs-efficient for majority of variables, if the entire mean squared error is studied. Of course, specific parameters of certain study can contribute to different evaluation conclusions for this procedure.
Year of publication2004
Bibliographic typeConferences, workshops, tutorials, presentations

Web survey bibliography - Vehovar, V. (139)