Accounting For Interviewer Variability In Small Area Estimation

摘要

Variation in interviewers' conduct could impact the data and become a major source of measurement error in interviewer administered surveys. This important source of variability has been largely overlooked in the small area literature. In this paper we develop alternative small area models that incorporate interviewer variability. We then assess the impact of interviewer variability on the inflation in the mean squared error (MSE) of the best predictor. A Monte Carlo simulation study is undertaken to assess the impact on the empirical best linear unbiased estimators.