Analysis of experiments embedded in probability samples
This paper develops a design-based inference mode for the analysis of randomized experiments embedded in probability samples .
To generalize results obtained with field experiments to large populations it is necessary to select the experimental units with a probability sample drawn from the intended target population. In addition, the analysis including tests about hypotheses of treatment effects must be based on (approximately) design-unbiased point and variance estimators. Such an inference procedures is proposed in this paper for a large set of experimental designs embedded in generally complex sample designs.