Figure 2.

Effect estimates of infrasound and traffic on measures of sleep quality over 3 nights in the laboratory-based study of the three-arm crossover study of 72 h of exposure to simulated wind turbine infrasound, sham infrasound, and traffic noise. Infrasound is represented in blue squares, sham infrasound in red triangles, and traffic noise in black circles. Error bars are 95% CI. Effects estimates are derived from mixed models of repeated measures where the participants were classed as random effects and exposure (3 levels), the order the exposure was received (1, 2, 3), the night of exposure (1, 2, 3), and interactions between the exposure by night and night by order as fixed effects. The least squares means procedure was used to address missing data. The exact numerical values for the estimated means and 95% CIs can be found in Table S3. (A) Sleep onset latency is the amount of time taken to fall asleep [2.48 min difference between infrasound and sham (95% CI: $-2.87$, $7.82$, $p=0.36$); Table 2]. (B) Number of sleep stage shifts is a measure of sleep stability [$-3.5$ shifts difference between infrasound and sham (95% CI: $-9.5$, $2.4$, $p=0.24$)]. (C) Proportions of the sleep period scored as each of the traditional sleep stages plotted together to test whether infrasound causes perturbation to sleep depth. Numerical values above each plot are the $p$-values for the difference between infrasound and sham infrasound. $p$-Values above the stacked columns are comparing infrasound to sham infrasound. (D) Arousal index is the number of cortical arousals detected during each hour of sleep as a measure of sleep quality [0.29 events/h difference between infrasound and sham (95% CI: $-0.58$, 1.16, $p=0.51$)]. Note: %, percentage; 1, non-REM sleep stage 1; 2, non-REM sleep stage 2; 3, non-REM sleep stage 3; CI, confidence interval; REM, rapid eye movement stage; SEM, standard error of the mean; W, wake.