ABSTRACT
After the survey data have been collected and all the essential steps of data processing (data entry, coding, editing, etc.) have been completed, a critical step must be implemented before the data can be analyzed. The survey data must be appropriately weighted. The weighting process essentially involves creating a new variable, say wi , for each respondent (labeled i) in the sample that will be referred to as the weight associated with the respondent. The weight can be interpreted as the number of individuals in the target population represented by the sample respondent. As an example, a weight of 100 indicates that the respondent represents himself/herself and 99 other persons in the target population. Except in special cases, wi ≥ 1 for all respondents, because at a
minimum, a respondent represents himself/herself, and wi =0 for all
nonrespondents. In most practical situations, the wi are not all equal even when all the sample members were selected with equal probability. This is due to so-called post-survey weight adjustments which attempt to reduce the standard errors of the estimates and/or compensate for the effects on the estimates of survey nonresponse and frame noncoverage. These adjustments allocate additional weight to some survey respondents who are selected to represent persons missed due to an incomplete frame or nonresponse. Therefore, even though the sample may be selected with equal probability, the weights assigned to the survey respondents can vary considerably.
