All Times ET
Program is Subject to Change
Estimation Approaches in Establishment Surveys (308405)*Wesley Yung, Statistics Canada
The availability of auxiliary information from registers gives considerable flexibility in estimation approaches in establishment surveys, and we summarise the uses of simple and more complex versions of calibration estimation, and the properties that make it useful in multipurpose surveys. We consider methods for dealing with outlying and unusual observations with large effects on estimates, and the ways in which bias and variance are traded off in these estimates. New forms of data can also be useful in estimation, and we summarise how this type of information can be used in combination with survey data.
The design-based approach is the most frequently used in establishment surveys, but other approaches are potentially beneficial. We consider model-based and Bayesian estimation approaches and when they might be most useful. As a specific example we take model-based small area estimation for business surveys, which is challenging because many of the assumptions of the multilevel models used as the basis of small area models are not met. We consider both area level and unit level models, and explain how these may also be made robust to outlying observations.