Selection of the Second-Stage Sample: Subareas Within Primary Sampling Units

Unlike the 1996 NHSDA, which used virtually the same segments surveyed in the 1995 NHSDA, the 1997 NHSDA basically surveyed a new segment sample. Approximately 95% of the 1997 sample, or 1,844segments, consisted of the previously unused units of the pairwise segment sample that was selected at the same time the 1995 and 1996 segment samples were selected. The remaining 5%, or 96 segments, overlapped with the 1996 survey year.

The 1997 second-stage sampling frame was based on the 1990 Decennial Census. Census blocks were combined with adjacent blocks based on the block identification number to create nonoverlapping area segments with at least 90 occupied dwelling units. The sample segment size allocations, based on optimal allocation and the expected precision requirements of individual strata, are given in Table D.1 for each stratum. These allocations assumed that about nine interviews would be completed per sample segment.

The area segments from each stratum of each of the PSUs were selected with probabilities proportional to a size measure. For each stratum, a composite size measure was defined to equal a weighted sum of the numbers of all dwelling units with weights proportional to the desired racial/ethnic sampling rates. Segments within the national Strata 1 to 5 were sorted to implicitly stratify the sampling frame to reduce sampling variability and to control the distribution of the sample. Segments in Strata 1 and 5 were sorted by metropolitan size, SU number, and socioeconomic status (SES) percentile based on a composite rental value index. Segments in Stratum 2 were sorted by metropolitan size, SU number, percentage Hispanic, percentage black, and SES percentile. Segments in Strata 3 and 4 were sorted by metropolitan size, SU number, percentage black, and SES percentile. Chromy's (1979) sequential selection routine was used to select the segments from the sorted segments within each stratum. A variable number of segments was selected from each SU in Strata 1 to 5 based upon each SU's composite size measure.

Prior to selecting segments from the national initial noncertainty stratum, the PSUs were divided into two groups: (a) high minority PSUs with 50% or more combined Hispanic and black eligible population, and (b) PSUs that had fewer than 50% minority population. Segments in high minority PSUs were sorted by the complement of the percentage Hispanic or the complement of the percentage black. The complement percentages were used to make sort transition more smoothly from high minority to low minority. Segments in low minority PSUs were ordered by SES percentile. Chromy's (1979) sequential selection procedure was used to select segments from the sorted list, selecting 12 segments from each SU with probability proportional to the composite size measure.

In the Arizona and California State supplements, a slightly different sorting of the sampling frame occurred to control the distribution of the sample. Segments from noncertainty PSUs were first ordered by percentage white, and then by SES percentile. California noncertainty SU segments were ordered first by a rural/urban indicator, then by SES percentile. Chromy's (1979) sequential selection procedure was used to select nine area segments from each SU with probability proportional to their composite size measure. There were no additional segments selected from the certainty PSUs because the certainty PSUs for both States were well represented in the national study component.

After area segments were selected from within the initial noncertainty stratum, sample segments were poststratified to allow for oversampling of concentrated minority population segments at the dwelling unit and person sampling stages. This poststratification of noncertainty segments contributed to the cost efficiency of the sample design by ultimately reducing the screening sample.

This cost efficiency was achieved by noting that the size of the screening sample could be reduced significantly by partitioning the initial noncertainty stratum into high racial/ethnic concentration strata. The following three noncertainty strata (numbered 6, 7, and 8, respectively) were created: high black, high white, and the remainder of noncertainty segments. These strata were defined based on dwelling unit and target population percentages of a segment as determined by the 1990 Census block data. The third-stage sample of dwelling units was allocated to the strata of Table D.1, and separate within-household person sampling probabilities were set by these strata to meet the design precision requirements at minimum cost.