Fig 1.
Number of comfort ratings by comfort state.
Black lines indicate the mean values. The mean comfort score for quiet states (N = 445, M = 4.2, SD = 0.8) is higher than that for noisy states (N = 445, M = 2.8, SD = 1.1).
Fig 2.
Distribution of ΔLAeq = LAeq[quiet]—LAeq[noisy] for different values of Δcomfort (difference of subjective comfort ratings between quiet and noisy states felt by each participant).
Black lines indicate median values. A linear model yields ΔLAeq/dB (A) = −1.54−5.91×Δcomfort (F(1,443) = 214, p < 0.001, R2 = 0.325).
Fig 3.
Dispersion of calibrated LAeq for the quietest background.
The values were obtained by adding to the raw LAeq the gains reported in S2 Table in S1 File.
Fig 4.
Distribution of LAeq for noisy and quiet states.
Thick black lines show median values. A few outliers are visible on the right side of the quiet state distribution. The outliers below 26 dB were considered due to inappropriate calibration and were removed from the dataset.
Fig 5.
Calibrated LAeq by comfort rating.
Thick black lines indicate the median values; the box limits the first and third quartile.
Fig 6.
Cumulative fraction of records reporting high comfort levels (4 or 5) as a function of the corresponding measured LAeq.
Respectively 82% and 69% of the records refer to sound levels below thresholds of annoyance as established by Italian regulations (i.e. 50 dB(A) during day-time and 40 dB(A) at night). Error bars indicate 95% confidence intervals, which are [76%, 87%] for day-time and [56%, 81%] for night-time data. Sample sizes are N = 211 (daytime) and N = 51 (nighttime).
Fig 7.
Distribution of LAeq in low and high comfort states.
The color code indicates the sources in each 3 dB(A) bin.