Multilevel modeling and regression as applied to community noise annoyance surveys

Tolerance to noise varies between communities, and between individuals comprising the communities. A multilevel modeling approach is useful for capturing such individual- and community-level variations. Here, we consider a model in which the community-level variations are sampled explicitly with community noise surveys, while the individual-level variations are sampled only in a statistical sense (i.e., hidden). The community-level variations are specifically quantified by the community tolerance level (CTL) [Fidell et al., J. Acoust. Soc. Am. 130(2), 791-806 (2011)]. Simulations based on the multilevel model indicate that the community- and individual-level variations have distinct statistical signatures, both of which are evident in noise annoyance surveys involving transportation noise. The annoyance curve for a previously unsurveyed community depends on the mean and variance of the CTL, and the sum of the hidden variances in noise tolerance and exposure among individuals in the community. Regression a...