A Subjective Bayesian Response BIAS Model

Abstract

In order to study the reponse bias in sample survey, we consider a simple response bias model via a subjective Bayesian approach (see Ericson (1969) ). It is possible to assess the magnitude of the response bias by the method of subsampling. One way is to measure a portion of the subjects twice. For instance, at a certain time period t, a first sample is drawn. At another time period t^*, a second sample is drawn such that the two samples are partially (or completely) overlapped. We try to obtain the responses of each of the overlapped elements in a way such that one of the measurements could be more precise (and hence more expensive) than the other measurement. Heneo the magnitude of the response bias can be measured. In this paper, we construct a simple response bias model. Optimal designs and estimates can be obtained via a subjective Bayesian approach.