Survival Modeling of Cumulative MLB Season Audience

• Activity Number: 649
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Ekaterina Sotiris*
• Keywords: survival modeling ; survey data ; marketing

Abstract

The Major League Baseball season runs March through September. Each of the 162 games is broadcast over the radio. Nielsen estimates the individual game audiences using electronic measurement in rolling panels. Every day, the number of panelists who wore their meter that day is assigned a sample weight, and their radio exposure data is recorded.

Typically, individuals in Nielsen's sample are monitored for shorter time periods than the duration of the baseball season. This poses a problem when we estimate the cumulative audience to the baseball season, as they may have listened to a game outside of their panel tenure. Observations on such persons are censored, i.e. only partially known. We propose using survival analysis techniques to utilize information on censored individuals when calculating audience estimates. Since exposure to a baseball program is a recurring event, repeated events survival analysis methodology is considered. We investigate the integral assumption that the censoring mechanism is not correlated with game listening.