ABSTRACT
This chapter discusses experimental methods aimed at quantifying non-sampling errors in different data collection modes, especially in the case of interactions between multiple changes in the survey design. Discussed are embedded experiments and a design-based inference approach for its analysis, time series methods for repeated surveys, and their combinations thereof. The design of the experiments, as well as analysis, is explained, including power analyses and optimal sampling design. The methods are illustrated with the redesign of the Dutch Crime Victimization Survey, where both a new data collection approach and a new questionnaire were introduced.
