|
Abstract:
|
In area probability samples, the primary sampling units (PSUs) are often counties or groups of contiguous counties. For household area samples, the secondary sampling units (SSUs, or segments) are often constructed using Census blocks, block groups, or tracts. Algorithms to form segments by grouping areas that are adjacent in numbering schemes (e.g., adjacent block numbers within tract, in a sorted list) sometimes result in discontiguous segments. This may be a problem if area measures (e.g., environmental data) are to be obtained or if multilevel modeling is to be used for analysis. In this paper we describe a study that motivated the development of an algorithm to optimize the configuration of segments based on Census blocks. The algorithm is described and evaluated, and ideas for enhancements are discussed.
|
