Strategy for Use of Found Samples in a Rigorous Monitoring Design

摘要

In the context of a probability sample, data from found samples have a high potential of misinference. Careful attention to representativeness is required to provide assurance that the found data are coherent with the probability sample data. A strategy of establishing representativeness by a process of similarity analyses is proposed, with similarity perceived relative to the use to which the found data are to be put. Two cases are developed. In one, the found sample is structured as a quasi-probability sample, and combined with the probability sample. In the other, the found sample is structured as a quasi-double sample, and used in the manner of a double sample of the probability sample.