Abstract:
|
In face-to-face survey data collections, interviewer observations recording selected features of sampled units represent a promising, cost-effective source of paradata that can be used for nonresponse adjustment. Recent research indicates interviewers can successfully record features of sampled units correlated with key survey measures and response propensity. However, these observations can be error-prone, limiting the effectiveness of nonresponse adjustment. Furthermore, no existing study has identified additional sources of interviewer variance of observation accuracy. This study performed analyses of justifications for interviewer observations in the National Survey of Family Growth (NSFG) on two key features of all sampled households: the presence of children under age 15 and expected probability of household response. Cluster analyses suggest that unique subgroups of NSFG interviewers do exist based on the observational strategies used. Multilevel models comparing the accuracy of the observations among the identified interviewer subgroups suggest that alternative strategies employed by the interviewers are more or less effective for improving the quality of the observations.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.