Selection of Primary Sampling Units

The 1997 national edition of the NHSDA used the same 115 PSUs selected for the 1995 and 1996 NHSDAs, plus a total of 18 supplemental PSUs from Arizona and California. These PSUs were selected to represent the Nation's total eligible population, including areas of high Hispanic concentration. The PSUs were defined as metropolitan areas, counties, groups of counties, Census tracts, and independent cities. The 115 PSUs in the national study comprised 43 certainty PSUs and 72 noncertainty PSUs. The 43 certainty PSUs were metropolitan areas with high Hispanic concentration that had been included in the NHSDA with certainty since 1988. The 72 noncertainty PSUs were selected with probability proportional to size (PPS) and minimal replacement to represent the balance of the Nation outside the 43 certainty PSUs. The national sample was supplemented by a PPS selection of 14 noncertainty PSUs from Arizona plus four noncertainty PSUs from California. Because the national sample provided representation for certainty PSUs in each State, no additional certainty PSUs were added to either sample.

The segments that formed the 43 certainty PSUs were partitioned into five strata:



• Stratum 1: Certainty-high-concentration Hispanic area segments. These segments had 71% or more Hispanics according to the 1990 Census.

 
• Stratum 2: Certainty-moderate-concentration Hispanic area segments. These segments had between 20% and 71% Hispanics according to the 1990 Census.

 
• Stratum 3: Certainty-low-concentration Hispanic area segments. These segments had fewer than 20% Hispanics but more than 2% of the combined Hispanic and black population according to the 1990 Census, and furthermore included fewer than 70% of the black population according to the 1990 Census.33

 
• Stratum 4: Certainty-high-concentration black area segments. These segments had fewer than 20% Hispanics but more than 2% of the combined Hispanic and black population according to the 1990 Census, and furthermore included at least 70% of the black population according to the 1990 Census.
  
  
• Stratum 5: Certainty-high-concentration white area segments. These segments had 2% or fewer of the combined Hispanic and black population according to the 1990 Census.
  
  
To complete the stratum definitions for the remainder of the national study, an initial global noncertainty stratum was created for the residual portion of the United States and defined as the initial noncertainty stratum representing the balance of the Nation.


At the final two stages of within-segment dwelling unit person sampling, segments in the noncertainty stratum for the national study were later partitioned into five substrata to improve the efficiency of the design by reducing the size of the screening sample.

Prior to the formation of the final strata, an optimal allocation procedure was used to allocate the sample to the six strata listed above. The results of the allocation suggested that total survey costs would be minimized for fixed precision when Hispanics were oversampled in the national study strata in which they were concentrated. Once sufficient screening interviews were conducted to identify the required number of Hispanic dwellings, more than enough black and white dwellings would be identified so that these interviews could be proportionally allocated to strata in the national study. For the Arizona/California supplement, there was no oversampling of Hispanics because each State's Hispanic allocation in the national study component was large enough to satisfy State-level precision requirements. Consequently, the three racial/ethnic groups were sampled in the supplement so that the combined national study and supplemental samples would result in a proportionally allocated sample.

The 43 certainty PSUs for the national study contained about 80% of the Hispanic population in the United States, according to the 1990 Census. They also contained approximately 50% of the blacks and about 40% of the whites in the United States. The segments of Stratum 1 had about 81% Hispanic dwellings on average and contained about 24% of the U.S. Hispanic population. The segments of Stratum 2 had about 32% Hispanic dwelling units and contained about 35% of the U.S. Hispanic population. The segments of Stratum 3 had about 5% Hispanic dwelling units and contained about 18% of the U.S. Hispanic population.

The remaining 72 noncertainty PSUs for the national study were selected from Stratum 6. The noncertainty PSUs, single counties with one exception, were selected with probability proportional to a composite size measure. The composite size measure was defined as the sum of racial/ethnic group dwelling counts weighted by the specified racial/ethnic sampling rates. This selection scheme allowed for targeting particular racial/ethnic subpopulation sample sizes. Chromy's (1979) probability minimum replacement sequential sampling scheme was used to select these 72 PSUs with probabilities proportional to their composite size measures. The use of a composite size measure ensured (a) roughly equal sample sizes per sample subarea (the second-stage unit of selection) and (b) roughly equal probabilities of selecting eligible individuals (the fifth-or final-stage unit of selection) within race/ethnicity, and age group.

A State-supplementary sample large enough to provide reliable estimates for Arizona and California was selected to participate in quarters 2 through 4 of the 1997 NHSDA. The certainty PSUs included in the supplement also were included in the national study. This supplement consisted of no additional certainty PSUs because a representative certainty-SU sample from both States was already participating in the national study.

The same first-stage sampling frame used for the noncertainty national sample also was the basis for selecting an independent supplement of 14 PSUs from Arizona and 4 PSUs from California. Because the noncertainty counties in Arizona were large geographically relative to population density, the counties representing the residual outside the certainty counties in Arizona were subdivided into smaller areas (Census tracts). Groups of these tracts satisfying a minimum population size requirement were defined as PSUs so that data collection would be more manageable. For the noncertainty California SU sample, no subdivision of counties was necessary; hence, each noncertainty-county comprised a single SU.

33 In the interest of readability for this report, "white" is used to indicate "white, non-Hispanic," and "black" to indicate "black, non-Hispanic."

This page was last updated on December 30, 2008.