Testing for the significance of changes in television ratings

Abstract

South African television ratings are obtained from the AMPS meter panel. This panel must be viewed as a complex non random sample. For such samples the effective sample size differs from the actual sample size. It has been found that, when equal weights are assigned to strata, the most reliable estimate for effective sample size can be obtained by considering every household as a sample cluster. This estimate of effective sample size can be incorporated directly into a test for significant rating change. For convenience this test is implemented graphically.