ABSTRACT
This chapter describes what tactics and techniques may be used to decompose effects that result from the choice of mode(s) into its two main components: mode-specific selection bias and mode-specific measurement bias. An overview is given of various methods to separate selection effects from measurement effects in empirical studies, and the weighting or regression-based inference methods to control for selection effects. These methods require strongly related auxiliary information. Standard available auxiliary information generally concerns socio-demographic variables, which are only weakly related to the target variables of a survey. To overcome this problem repeated measurement experiments and re-interview designs are discussed with the purpose of collecting strongly related auxiliary information in the re-interview. This information is used to construct regression estimators that can better correct selection bias. The chapter also provides a cost-benefit analysis of experimental designs focused on such decompositions.
