|Page tools: Print Page Print All|
WEIGHTING, BENCHMARKS AND ESTIMATION PROCEDURES
Weighting is the process of adjusting results from a sample survey to infer results for the total in-scope population. To do this, a 'weight' is allocated to each sample unit; that is, a person or a household. The weight is a value which indicates how many population units are represented by the sample unit. Separate person weights were developed for each survey in the Australian Aboriginal and Torres Strait Islander Health Survey (AATSIHS) and separate household weights were developed for the National Aboriginal and Torres Strait Islander Health Survey (NATSIHS) as outlined below.
The first step in calculating weights for each person or household is to assign an initial weight, which is equal to the inverse of the probability of being selected in the survey. For example, if the probability of being selected in the survey was 1 in 45, then the person would have an initial weight of 45 (that is, they would represent 45 people).
After calculating the initial person weights, an adjustment was incorporated into the weighting to account for Indigenous persons not covered by the sample. The initial household weights (NATSIHS) were also similarly adjusted.
Person and household weights are calibrated to independent estimates of the population of interest, referred to as 'benchmarks'. Weights calibrated against population benchmarks ensure that the survey estimates conform to independently estimated distributions of the population rather than to the distribution within the sample itself. Calibration to benchmarks helps to compensate for over- or under-enumeration of particular categories of persons and households, which may occur due to the random nature of sampling, non-response, non-identification or various other undercoverage factors. This process can reduce the sampling error of estimates and may reduce the level of undercoverage bias. For more information on undercoverage, see Data quality page of this Users’ Guide.
All person weights were benchmarked to the Australian Aboriginal and Torres Strait Islander estimated resident population living in private dwellings of Australia at 30 June 2011, based on the 2011 Census of Population and Housing. The benchmarks and hence the estimates from the survey do not (and are not intended to) match estimates of the total Australian Aboriginal and Torres Strait Islander estimated resident population (which include persons in non-private dwellings) obtained from other sources.
A standard approach in ABS household surveys is to calibrate to population benchmarks by state, part of state, age and sex. In terms of the effectiveness of 'correcting' for potential undercoverage bias, it is assumed that the characteristics being measured by the survey for the covered population are similar to the uncovered population within weighting classes, as determined by the benchmarking strategy. Where this assumption does not hold, biased estimates may result.
The ABS does not produce Indigenous dwelling counts. Therefore, for the 2012-13 NATSIHS, household level benchmarks were not available. Instead, the household weights for this survey were derived from the person level benchmarks. This was done by assigning the initial household weight (after the adjustment) to all Indigenous persons in the household. These weights were then calibrated to the person level benchmarks with the constraint that each person in the household must have the same final calibrated weight. The resulting weight was assigned as the final household weight. That is, the weights were calibrated in such a way that the household weights will reproduce the number of known person level benchmarks if all people in a household are included. Therefore, the sum of the household weights will only provide an estimate of the number of Indigenous households. This method was then analysed to ensure that person and household level estimates were as consistent as possible.
Note that household weights are only available on the NATSIHS data files, and not on the NATSINPAS or the Core data files.
Two types of person weights were produced for the 2012-13 NATSIHS, NATSINPAS and the Core survey files:
For person level weights, 250 replicate weights were produced to ensure accurate estimates of the standard errors. The age groups (years) available for use in calibration were: 0-1 years, 2-4 years, 5-9 years, 10-14 years, 15-17 years, 18-19 years, 20-24 years, 25-29 years, ... 60-64 years, 65 years and over.
The person weights were simultaneously calibrated to the following population benchmarks:
NATSIHS and the Core files person weights:
NATSINPAS file person weight:
The biomedical person weights were simultaneously calibrated to the following adult population benchmarks:
NATSIHS file biomedical person weight:
NATSINPAS file biomedical person weight:
The Core file biomedical person weight: