APPENDIX A Notes on Weighting: 1987 AIDS Public Use Data Documentation The final case weight in data locs. 190-195 produces national estimates based on the full 22 weeks of data included in the 1987 AIDS Knowledge and Attitudes data, i.e., for data collected from August 10, 1987 through the end of the survey year. The estimator embodied in the final case weight includes a ratio-adjustment to benchmark population estimates by age, sex, and race. This adjustment also attempts to reduce bias resulting from nonresponse and household noncoverage. The final case weight must be adjusted if the data user intends to develop estimates for shorter time periods, e.g., to analyze data for specific calendar months. The data contained on this data will not provide reliable estimates for time periods of less than 4 weeks. To develop estimates for subsets of the 1987 AIDS data data, e.g., for specific calendar months, the data user must first identify cases by date of interview. This can be done by selecting on two fields: the processing quarter (data loc. 5) and the processing week code (data locs. 19-20), which is numbered from 01 to 13 within each quarter. The table below shows how the quarter and processing week codes correspond to actual interview weeks. For example, to calculate estimates of AIDS knowledge for November 1987, the user would select only those cases where processing quarter = 4 and where processing week code = 04-07. Processing Processing Interview Quarter Week Week 3 05 Aug. 10 - Aug. 16 3 06 Aug. 17 - Aug. 23 3 07 Aug. 24 - Aug. 30 . 08 Aug. 31 - Sept. 6 . 09 Sept. 7 - Sept.13 . 10 Sept.14 - Sept.20 . 11 Sept.21 - Sept.27 . 12 Sept.28 - Oct. 4 3 13 Oct. 5 - Oct. 11 4 01 Oct. 12 - Oct. 18 4 02 Oct. 19 - Oct. 25 4 03 Oct. 26 - Nov. 1 . 04 Nov. 2 - Nov. 8 . 05 Nov. 9 - Nov. 15 . 06 Nov. 16 - Nov. 22 . 07 Nov. 23 - Nov. 29 . 08 Nov. 30 - Dec. 6 . 09 Dec. 7 - Dec. 13 - 1 - Processing Processing Interview Quarter Week Week . 10 Dec. 14 - Dec. 20 . 11 Dec. 21 - Dec. 27 . 12 Dec. 28 - Jan. 3 4 13 Jan. 4 - Jan. 10 Final weights for subsets of the 1987 AIDS Knowledge and Attitudes data, such as for a single month of data, may be calculated by multiplying the final case weight in data locs. 190-195 by 22/n, where n is the number of weeks of data in the subset being examined. The resulting adjusted final case weight will be approximate since it does not directly incorporate the age/race/sex poststratification adjustment described earlier. These weights for an n-week period may not yield estimates consistent with the benchmark population estimates by age, sex, and race. Data users who wish to develop more accurate estimates for subsets of the data, e.g., for specific months, can customize their own age/sex/race poststratification procedures if desired: 1) Create a temporary adjusted final case weight by multiplying the interim case weight (data locs. 172-177) by 22/n. 2) Determine age/sex/race poststrata appropriate to the intended analysis, making sure that each poststratum (e.g., black men aged 18-29) contains at least 50 cases. 3) Sum the temporary adjusted case weights (i.e., the interim weights times 22/n) within each poststratum and divide the benchmark population in the poststratum, e.g., a Census count or estimate, by this sum. The result of this division is the age/sex/race poststratification adjustment factor. 4) Multiply the temporary adjusted case weight by the appropriate age/sex/race poststratification factor to obtain final adjusted case weight. - 2 -