Simulation and Application of Bayesian Dose Uncertainty Modeling for Low-Boom Community Noise Surveys
In dose-response modeling, failing to account for dose uncertainty can cause artificial flattening of the estimated slope of the dose-response curve. Previous analyses of NASA sonic boom community noise survey data utilized a Bayesian multilevel logistic regression model, which did not account for dose uncertainty. The current work extends the model to account for either classical or Berkson dose uncertainty. The extended model is applied to two simulated dose-response datasets to illustrate conditions under which the dose uncertainty term does and does not correct for the artificial flattening introduced by dose uncertainty. Finally, the extended model is applied to two previous NASA sonic boom community noise surveys. The resulting dose-response curve slope for the average participant is 5 to 10% steeper, but the difference in the noise dose that elicits a 5% highly annoyed response is small (less than 1 dB). The difference remains insignificant when producing population summary dose-response curves. Commentary is included on applicability to future X-59 low-boom community noise survey data modeling and analysis.