Local market television measurement: A comparison of the statistical reliability of three methodologies

Date of publication: June 15, 1992

Company: Nielsen


Since the introduction of the People Meters in the U.S., attention has been centered on its goodness as an accurate measurement method with particular emphasis on validation studies of audience levels. Relatively little attention has been focused on comparing the statistical reliability of the ScanAmerica® People Meter service with that offered by the traditional methodologies it is replacing. This paper offers such a comparison. It notes that statistical reliability evaluations require more than just simple comparisons of raw in-tab sample sizes. The three audience measurement methods are based upon different sample design features, and these have a substantial impact on statistical reliability. Through the use of statistical replication studies and Arbitron standard error models, the statistical efficiency values associated with each of the three methods are derived and compared. This statistic summarizes in a single quantity the total impact on statistical reliability that results from the various sample design features associated with each methodology. To measure the combined value of both sample size and the statistical efficiency factor, a quantity called Effective Sample Base (ESB) is also calculated. For purposes of this paper, the ESB statistic will serve as the "bottom line," or final arbitrator, for comparing statistical reliability across the three methodologies. The conclusion from these data analyses is that the ScanAmerica® People Meter service offers an improvement in reliability compared to both of the traditional methodologies it is replacing.

  • PDF
  • This could also be of interest