A comparison of predictors for a measurement error Fay Herriot model
Conference
64th ISI World Statistics Congress - Ottawa, Canada
Format: IPS Abstract
Keywords: jackknife, prediction, survey
Session: IPS 64 - Measurement Error Modeling: Advances and Applications
Tuesday 18 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
The traditional approach in small area estimation is to use population-level auxiliary information from censuses or administrative databases. In practice, use of auxiliary information from surveys is often pragmatic. When the auxiliary variable is from a sample survey, the covariate is contaminated with measurement error. Most procedures for small area estimation with measurement error in the covariate assume that the covariate and response are independent. When the response and covariate are from the same survey, this assumption of independence is unreasonable. We compare the properties of several small area estimation procedures under the assumption that the measurement error in the covariate is correlated with the sampling error in the response.