Abstract:
|
Nonresponse bias is an ongoing concern to all who collect data via surveys. It occurs when nonresponding sampled units differ from those who respond, in ways related to the survey outcome variables of interest. Although every effort should be made to minimize nonresponse at the point of data collection to minimize the risk of bias, these efforts are rarely adequate in preventing a high degree of nonresponse. Thus a post-hoc analysis of nonresponse is often necessary to determine the effects of nonresponse on survey outcomes. In this study, we will analyze nonresponse in a sample of 2,067 urban households selected for a survey of community health needs in Chicago. As the study was conducted via face to face interviews, we gathered neighborhood (block) level data and observable unit level data. In addition, we appended block group level Census Planning Data to the block groups to which the sampled blocks belong. We have used these data to model total response, contact and cooperation propensity among sampled households. Those variables predictive of these various response propensities were examined for their association with survey outcome variables of interest, including chronic and
|