Journal of Official Statistics, Vol.17, No.2, 2001. pp. 149–161
Estimation of Interviewer Effects on Multivariate Binary Responses in a Community Based Survey
Abstract:Estimation of interviewer effect in a survey is important in providing means to control nonsampling errors. It is often desirable to estimate the overall interviewer effect when multiple responses are obtained by one interviewer, especially in surveys where the interviewers' subjective judgment is used in determining the outcomes of the response. In this article I propose a two-stage logistic normal model for estimating interviewer effects on multivariate binary responses from surveys. The model is an extension to Anderson and Aitkin (1985) in univariate binary cases. A maximum likelihood estimation method using Gauss-quadrature points are presented with discussions on computational aspects of the approach. Data from a community based dementia survey are used to motivate and illustrate the proposed approach.
Keywords:Multivariate binary outcomes; logistic normal model; Gauss quadrature.
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group