Online Program

Weighting and Sample Matching Techniques for Reducing Bias in Online Convenience Panels (302753)

Keywords: sampling, sample matching, data fusion, convenience panel, bias, weighting, rim weighting

Using an online convenience panel is an economical and quick way of getting research data. While such panels can address topics ranging from consumer purchase behavior to voting intentions, can we rely on the results? Although large, these panels may not reflect the behavior or sentiment of the population in which we are interested. In turn, results may not be accurate and therefore may not be actionable. How can such panels be best sampled to minimize bias? This presentation will examine a methodology to select and weight samples from a convenience panel to model a representative sample of the target population. We use a smaller, representatively drawn sample as a reference. A rim weighting procedure produces a demographically representative sample, and a data fusion process matches common demographic and online usage variables between the convenience and representative samples. The results are then verified by examining online behavioral metrics, such as reach and time spent variables, for top brands and channels.