Multiple imputation for measurement error correction in survey data

In sample surveys, the uncertainty of parameter estimates comes from two main sources: sampling and measuring the study units. Statistical quality is often described with terms of variance, bias and MSE. Some aspects of survey errors are quite well understood (e.g. sampling errors, nonresponse errors) and reported but others, like measurement errors, are often neglected. Reliable measurement is essential in surveys. However reliability of measurement is not yet an accepted standard for describing the quality of survey data. An interesting approach to adjustment for measurement errors is multiple imputation for measurement errors (MIME) (Cole et al., 2006; Padilla et al., 2009). In MIME approach measurement errors are treated as a missing data problem. Multiple Imputation (MI) (Little and Rubin, 2002) is used to impute the missing (latent) true scores. The aim of this study is to investigate the MIME approach and its applicability under different sampling designs. In order to assess the statistical quality of the estimates, a simulation study will be implemented. Artificial data will be generated based on Finnish ECHP data of years 1996 and 2000, where the variables of interest such as income are measured both by interview and by administrative registers.