DESCRIPTION OF SURVEY POPULATION 89 As discussed in Section 3.1, these two groups combined represented 7,018 persons-15.0 percent of the 46,905 persons enumerated in the final 105 survey sites. This represents a sizeable amount of missing data, and compensation was made by adjusting the base survey weights described previously in this section. The inverse of the probability of selection was multiplied by a factor meant to compensate partially for both of these two types of missing data simultaneously. The simplest scheme for the adjustment would be to multiply each base weight by a factor that corrects for the missing data rate. That is, each weight could be multiplied by the factor 1/(1 - 0.15) = 1/0.85 = 1.18, and the sum of weights for all examined persons would represent the sum of weights for all examined and unexamined persons in the survey sites. Such adjustment assumes, though, that the missing persons represent a random sample of the entire population-an unwarranted assumption. Since the age, sex, and site of each person who was not examined was actually known from the household register, an adjustment was made separately within age x sex x sampling strata sub- groups or cells. The cells were constructed separately within each of the 13 sampling strata by forming 17 age groups and cross- classifying them with sex. Cases that had missing data on age or sex were placed in a separate cell. Thus, 35 missing-data adjustment cells were constructed within each- stratum, and a total of 455 cells (35 x 13) were constructed for the entire sample. Within a cell, two weighted counts were obtained: the sum of base weights for all enumerated persons (including absentees and nonrespondents), and the sum of base weights for examined persons only. An adjustment factor was obtained by dividing the sum of weights for enumerated persons by the sum of weights for examined persons. Table 3-2 displays the 35 age-sex-specific adjustment factors computed for the first stratum (East Terai). For ex- ample, for males under one year of age in this stratum, the weighted sum for enumerated persons was 28,591.91 and for examined persons, 24,092.61. Thus, the age-sex-specific adjustment factor for males under one year in the East Terai was computed as 28,591.91/24,092.61 = 1.1867 or 1.19. All adjustment factors were rounded to three significant digits. In addition, if the adjustment factor for any one cell was larger than 2.00, the cell was collapsed with neighboring cells until an adjustment factor for the collapsed cell was less than 2.00. The collapsing proceeded first by age, collapsing to ten-year age groups within a sex group (e.g., ages 10-19, 20-29, etc.), and then by sex for ten-
