Multilevel modeling of recent community noise annoyance surveys

Abstract

Noise social surveys are typified by large variations in annoyance ratings for equivalent noise exposure. Some of this variation has been loosely linked to constructs that describe individual participants such as noise sensitivity. Multilevel statistical analysis provides a framework to study the relationships between annoyance and such constructs. The current effort applies multilevel analysis to two recent social surveys of impulse noise for which noise sensitivity was assessed indirectly via a baseline survey. For both surveys, multilevel models fit the data better than ordinary regression models. Multilevel models also quantify the variation in individual annoyance trends due to differences among participants. Noise sensitivity describes a limited proportion of the observed variation.