Abstract
A simple measurement error model is introduced for continuous data collected
by interviewers. The model makes a clear distinction between three different
sources of randomness, namely, samle selection, interviewer assignment,
and interviewing. The concept of interviewer variance is defined in the
context of this measurement error model, and the problem of estimating
the interviewer variance is considered, assuming simple random sampling.
A simulation study indicates that estimates of the interviewer variance
are unstable, especially when the interviewer variance is small (in which
case the effect of interviewer variability on the main survey results
may still be severe)
(1) Statistics
Sweden, S-10451 Stockholm and Department of Statistics, S-106 91, Stockholm.
(2) Stockholm
University, Department of Statistics, S-106 91, Stockholm.
|