Optimal Estimation of Interviewer Effects for Binary Response Variables through Partial Interpenetration

Estimates of sampling variance underestimate the variance of survey estimates when there are strong interviewer effects. However, interviewer effects are rarely considered in complex field surveys due to the high costs involved with adapting an interpenetrating design to produce estimates of them. This paper demonstrates how interviewer effects can be estimated by taking a sample of interviewers using the technique of partial interpenetration. The partial interpenetration approach is developed to produce cost-optimal survey designs for the estimation of the interviewer effect and population means for non-linear response variables.