Multilevel modeling and regression of community annoyance to transportation noise.

A multilevel (hierarchical) model is devised that separates noise tolerance into variations occurring at the levels of individual listeners and communities. This approach successfully describes the characteristics of real community transportation noise surveys, with the individual-and community-level variations producing distinct statistical signatures, both of which are evident in the surveys. Predictions are provided for quantities such as the probability of annoyance based on the observed noise level and statistical parameters characterizing the community tolerance. Regression analyses are performed using a multilevel, generalized linear model, which provides an appropriate generalization encompassing both no pooling (separate community-by-community analysis) and full pooling (all communities together) of survey data, and enables noise tolerances and their variations at the individual and community levels to be distinguished and quantified. Variations in individual tolerance and sound exposure within communities are found to be larger than variations in tolerance between communities; however, the variations between communities are still significant and observable. Analysis of multiple types of transportation noise with the multilevel model indicates that tolerance is highest for railway noise with low vibrations, followed by roadway noise, airport noise, and railway noise with high vibrations, as consistent with previous studies.