Rationale and Procedures for Weighting the SSI Study Data

Abstract

This paper describes the rationale and procedures for developing a set of weights for the multisite study of former recipients of Supplemental Security Income (SSI) for drug addiction and alcoholism (DA&A). The sample weights corrected for two factors: differing sampling rates among the sites and incomplete sampling frames and non-response at the sample-recruitment stage, which resulted in samples with demographics discrepant from those of the represented populations. We derived the first component of the weights by assigning a base weight calculated as the reciprocal of the selection probability for each case. The second component of the weights was calculated using an iterative raking procedure using known population marginal totals for age, gender, and race. Application of the resultant weights to the study sample yields an aggregate sample that more closely approximates the characteristics of the aggregate population represented by the nine study sites. When conducting analyses using the cross-site data, we recommend the use of a replication technique to obtain corrected variance estimates in conjunction with the sample weights.