1 / 24

Survey Weights: An Example from Honduras

Survey Weights: An Example from Honduras. Bob Gerzoff U.S. Centers for Disease Control and Prevention. 8 th CAMDI Workshop, Panamá City, 28–29 November 2006. Outline. Why weight? Steps to calculating weights. Sample calculation from Honduras. Why Weight?.

pete
Download Presentation

Survey Weights: An Example from Honduras

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Survey Weights:An Example from Honduras Bob Gerzoff U.S. Centers for Disease Control and Prevention 8th CAMDI Workshop, Panamá City, 28–29 November 2006

  2. Outline • Why weight? • Steps to calculating weights. • Sample calculation from Honduras

  3. Why Weight? • Simple averages assume everyone has the same chance of being selected for the survey. • Need to correct for unequal probability of selection. • Unequal selection probability impacts both the means and their confidence intervals

  4. Why Weight(2)? • Need to correct for non–response. • Want to adjust results to reflect the larger population from which the sample was taken

  5. Steps to Calculating Weights • Calculate a “sampling weight” • Adjust the sampling weight for non–response

  6. Calculating the Sampling Weight • Sampling weights are the inverse (reciprocal) of the probability of selection. • Sampling weights are calculated for each level of sampling. • Sampling weights for each level are multiplied together

  7. Sampling Weights in the Honduran Sample

  8. Sampling Weights in the Honduran Sample • Four of 20 sectors were chosen. • The probability of selection is 4/20. • The sampling weight for the sector level (for all sectors) is 20/4=5. • This weight applies to all individuals the sector.

  9. Sampling Weights in the Honduran Sample • In sector 2, there are ~17 compactos (blocks). • The probability of selecting any compacto is 3/17. • The sampling weight for the compacto level (for all compactos in sector 2 ) is 17/3 • This weight applies to all individuals in this compacto.

  10. Sampling Weights in the Honduran Sample • The sampling weight for all 144 individuals in sector 2 is: 5 * 17/3 = 28.3

  11. Sampling Weights in the Honduran Sample • The intent was to sample all households and all adults within each chosen sector/compacto. • The probability of selection at both those levels is therefore 1.

  12. Adjusting the Weights for Non-Response • Although the intent was to sample every household and all adults, not every or every adult was sampled. • Need to adjust the weights for non-response at the household and individual level.

  13. Adjusting the Weights for Household Non-Response

  14. Adjusting the Weights for Household Non-Response • In sector 2, compacto 11, there were 70 households: 50 were sampled. • Each household should therefore represent 70/50 or 1.4 households.

  15. Adjusting the Weights for Individual Level Non-Response • Adjustments to the sampling weights for individual level non–response are made specific for each age and sex combination.

  16. Adjusting the Weights for Individual Level Non-Response

  17. Adjusting the Weights for Individual Level Non-Response • In sector 2, compacto 11, there were 43 females age 20-39 to survey: 20 were surveyed. • Therefore, each female age 20-39 in sector 2, compacto 11 represents approximately 43/20 (2.5) individuals.

  18. Adjusting the Weights for Individual Level Non–Response

  19. Adjusting the Weights for Individual Level Non–Response • In sector 2, compacto 11, there were 4 males 65 or older to survey: Three were surveyed. • Therefore, each male 65 or older in sector 2, compacto 11 represents approximately 4/3 (1.3) individuals.

  20. Adjusting the Weights for Individual Level Non-Response • For a male, 65 or older, in sector 2, compacto 2 the sampling weight becomes: 28.3*1.3=36.8 • For a female, age 20-39 in sector 2, compacto 2, the sampling weight becomes: 28.3*2.5=70.8

  21. Scale the Weights • When all the weights have been calculated, they are scaled to represent the target population.

  22. Scale the Weights • If this is done to maintain the population subcategory relative sizes it is called, Raking.

  23. Standardization • The age/sex distribution of the sample may not be the same as the target population. • Raked weights automatically “correct” for this problem

  24. Standardization • If the age/sex distribution of the sample does not differ markedly from the sample, then we may decide not to standardize. • We may, however, decide to standardize so that our estimates reflect the larger target population.

More Related