A Ratings Reliability report provides you with information that’s important to understanding the role of Nielsen Audio’s sampling on your station rating estimates.
There’s more to a rating than simply the reported value. That reported value is only an estimate of the actual rating that exists for the entire market population.
Let’s say a station has an AQH rating of 0.5 for a given survey. That means the information from the sample for that survey indicates that, on average, .5 percent of persons in the market were exposed to the station for at least five or more minutes of listening (for Diary data) or three or more minutes of listening (for PPM data) during each 15-minute period. (For PPM data, prior to January 2025, stations qualified if they received credit for five or more minutes). But, that’s only part of the story.
The natural question is, what is this average for all people in the market? Perhaps the average for the market population is actually 0.4 or 0.7. The Ratings Reliability Estimator allows you to calculate measures — standard error and confidence interval — that help you determine a likely range on that population average.
Standard error is a statistical term for the amount of uncertainty involved in estimating a rating.
As with any estimates that come from a survey that uses a sample of the population, Nielsen’s radio listening estimates are subject to uncertainty. The uncertainty comes from the fact that any one of the possible random samples of the population will result in different estimates than any other. The variation in these hypothetical estimates over all possible samples is called sampling error. The standard error is a measure of this sampling error.
The confidence interval provides you with a range of plausible values for the actual population average. A confidence interval range of 0.4 to 0.7 for an AQH rating of 0.5 tells you that it is likely that the actual population rating is somewhere between 0.4 and 0.7. Some caution should be exercised, though, in interpreting confidence interval ranges. Confidence intervals by themselves don’t provide a complete picture of the likelihood of the population rating being any particular value. For instance, in this example, the most likely value of the population rating is 0.5 and the values near 0.5 are more likely values for the population rating than are the extremes of the confidence interval.
The confidence level tells you how strong the evidence is that the population rating is contained in the confidence interval range. The larger the confidence level, the larger the likelihood that the confidence interval range contains the population rating.