Abstract:
|
Nielsen regularly produces reliable survey design-based audience estimates of radio listening within the radio Metro markets in the US. This is accomplished in the larger markets using electronic measurement within radio panels, where sampled households report their listening for up to two years. However, audience estimates from these panels are sensitive to panel fluctuations and outlier households that may exhibit extreme listening. This may cause serious changes in the rankings of radio stations in a market between consecutive survey periods.In this study, we investigate the application of both design-based and Bayesian methods of ranking radio stations within each market. Stations are typically ranked using only the mean listening per time period, which ignores the reliability of each individual mean. We will assess the changes in ranking when the variances of the means are incorporated into the ranking and estimate the rankings for small demographic groups. We will utilize a model-based method, specifically small area estimation techniques, and compare the result to those of the typical design-based estimates.
|