How many ways is it worth slicing the social survey data cake?

Abstract

As when serving portions of cake, social survey findings on the prevalence of noise-induced annoyance may be sliced in many ways to create dosage–response relationships. Schultz (1978) divided ‘‘clustering’’ from ‘‘nonclustering’’ surveys; FICON (1992) sliced off a few inconvenient data points; several researchers and agencies have found reasons to develop separate relationships for certain noise sources; while yet others question the rationale for developing dosage–response relationships in the first place. Are the various dosage–response relationships usefully different from one another, and artfully enough sliced to support regulatory policymaking or theory development? Might differences among dosage-response relationships inferred by different means reflect statistical artifacts, or the influences of nonacoustic factors? This paper discusses applied and other uses for different slices of the same cake, and how many (more) slices might be useful.