Abstract:
|
For multiple and sound reasons, survey science is largely not an experimental science. Many surveys contain embedded small-scale experiments, for example, variations in mode, question wording, and incentives. Major questions, however, such as tradeoffs between cost and data quality, or between cost and decision quality, cannot be addressed experimentally. In this talk, I attempt to elucidate the extent to which a survey microsimulator--specifically, the World's Simplest Survey Microsimulator (WSSM)--can provide credible or actionable insight into the tradeoffs between survey costs and data quality. One advantage of the microsimulation approach is that the truth about the population is known, and multiple measures of data quality can be calculated. The most problematic aspect of microsimulations is, of course, validation. Results of several numerical experiments will be presented.
|