Design and analysis of experiments embedded in complex sample surveys

A sample survey has many potential sources of error that generally have a systematic effect on the outcomes. Therefore, many large-scale field experiments are conducted to optimise data collection strategies and questionnaires, for example. Such experiments are usually built into the sample design of the study.

In this thesis, design-based analysis methods are developed to test hypotheses about effects of different treatments (e.g. questionnaires or data collection methods) on sample survey estimates. The analysis methods take into account both the probability structure of the sample design by which experimental units are drawn from a finite population and the randomisation mechanism of the experimental design by which the experimental units are randomised across treatments. The developed methods are applied to different experiments conducted at Statistics Netherlands to improve survey processes.

Van den Brakel, J.A. (2001). Design and analysis of experiments embedded in complex sample surveys. Dissertation, Erasmus University Rotterdam.