|Page tools: Print Page Print All|
About this Release
Focusing on good estimates of the "underlying trend" of the series, and how it changes over time, could lead to quite different survey designs. Previous work suggests that a sample rotation pattern with no month to month overlap would provide better trend estimates. Unfortunately such a rotation pattern gives poor estimates of month to month change.
This paper considers an alternative estimator, the linear composite estimator, in combination with various sample rotation patterns. A rotation pattern is presented in which individuals are sampled for two successive months out of every four months, giving a 50 percent overlap of sample between successive months. By using composite estimation this rotation pattern yields improved estimates of trend while maintaining good estimates of month to month change.
These documents will be presented in a new window.