Consequences of improper analysis of probabilistically sampled data and suggestions for an alternate method of analysis

This study reviewed the purpose and practice of weighting, particularly as regards disproportionately sampled populations. Weighted data cannot be simply unweighted by multiplying by the reciprocal of the weights. Doing so produces erroneous results. When weights are used for the sole purpose of correcting bias due to oversampling the minority portion of a population containing a minority and a majority, proper analysis can be performed without the weights. The population proportion of a disproportionate sample can be restored by replicating the majority and minority portions of the data. These replications can then be concatenated to produce a dataset that may be used under the assumptions of simple random sampling. (Published By University of Alabama Libraries