Propensity-Weighted Web Survey

Because Web surveys are often convenience samples, traditional methods for statistical inference do not readily apply. Propensity scoring is one attempt to correct for the selection bias of a nonrandom Web sample. Broadly speaking, the propensity scoring adjustment is accomplished by reweighting the convenience sample such that the distribution of so-called propensity variables corresponds to that of a reference sample. Propensity variables [or propensity questions) can be any survey questions that have been answered both by the Web survey participants and by the respondents of the reference sample. The reference sample is a separate probability sample (e.g. a random-digit dialed [RDD] phone survey) from a possibly much shorter survey that contains only the propensity questions.