Alternative analysis of sleep-awakening data

Noise Control Engineering Journal

Volumn 55, Issue 2, 2007, Pages 224-245

  Anderson, Grant S ^a   Miller, Nicholas P ^a  

^a Harris Miller Miller and Hanson Inc   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

AIRCRAFT NOISE; AIRPORT OPERATIONS; SLEEP-AWAKENING;

ACOUSTIC NOISE; AIRCRAFT; DATA REDUCTION; SENSITIVITY ANALYSIS;

SLEEP RESEARCH;

EID: 34248585021     PISSN: 07362501     EISSN: None     Source Type: Journal    
DOI: 10.3397/1.2711617     Document Type: Article

References (42)

1. K. S. Pearsons et al., "Analyses of the predictability of noise-induced sleep disturbance," Report HSD-TR-89-029, Noise and Sonic Boom Impact Technology, Human Systems Division, Air Force Systems Command, Brooks Air Force Base, Texas (1989).
2. Federal Interagency Committee on Noise (FICON), "Federal agency review of selected airport noise analysis issues" (1992).
3. S. Fidell et al., "Noise-induced sleep disturbance in residential settings," Report AL/OE-TR-1994-0131, Occupational & Environmental Health Division, Armstrong Laboratory, Wright Patterson Air Force Base, Ohio (1994).
4. S. Fidell et al., "Field study of noise-induced sleep disturbance," J. Acoust. Soc. Am. 98(2) (1995).
5. Federal Interagency Committee on Aviation Noise (FICAN), "Effects of aviation noise on awakenings from sleep," www.fican.org/pages/sleepdst. html(1997).
6. S. Fidell et al., "Effects on sleep disturbance of changes in aircraft noise near three airports," J. Acoust. Soc. Am. 107(5) (2000).
7. S. Fidell et al., "Noise-induced sleep disturbance in residences near two civil airports," NASA Contractor Report 198252, Contract NAS1-200101 (1995).
8. American National Standards Institute (ANSI), "Quantities and procedures for description and measurement of environmental sound - Part 6: Methods for estimation of awakenings associated with aircraft noise events heard in homes," ANSI S12.9-2000/Part 6 (2000).
9. Gezondheidsraad, Health Council of the Netherlands, "The influence of night-time noise on sleep and health," Report U 1007/WP/718-K, The Hague (2004).
10. European Parliament and Council Directive 2002/49/EC, Annex I, OJ: 189, 18.7.2002, p. 12 (2002).
11. H. M. E. Miedema, W. Passchier-Vermeer and H. Vos, "Elements for a position paper on night-time transportation noise and sleep disturbance," TNO Inro report 2002-59 (2003).
12. Data from the original studies were provided to the current authors by Larry Finegold and Robert Lee - from records of the U.S. Air Force, sponsors of the original studies.
13. The expectation was for approximately the same number of aircraft-noise events in each of the two data sets, because the number of subjects was nearly the same. However, the random split resulted in one set having nearly double the number of aircraft-noise events as the other. That larger set was chosen for "regression," with the smaller set retained for "validation. "
14. The GLZ module within Statistica 7.1 (StatSoft, Inc., Tulsa, Oklahoma, www.statsoft.com) was used for all logistic regressions. Interaction terms were not investigated.
15. Other functional forms from the literature are simpler than this logistic form - for example, a linear regression. However, because the underlying data distribution is binomial, rather than Gaussian, least-squares regression is not proper for these data - certainly not to determine uncertainty bounds. The binomial requirement therefore undercuts the otherwise simple analysis for linear regression. In addition, it is highly desirable to have a functional form that is asymptotic to zero and unity, because it makes more realistic the multiple-aircraft joint probabilities, which can easily approach unity. For all these reasons, a functional form that transitions smoothly between zero and unity is highly desired. The logistic function is the simplest and most widely used of such transitional functions.
16. Indoor ambient sound level was dropped for two additional reasons: (1) this parameter was differently defined/measured at the three air facilities, and (2) values for it were missing from the database for many aircraft noise events.
17. Although this regression equation contains 32 sensitivity regression coefficients, these all collapse into just two parameters - their mean and standard deviation - when regression results are used. For that reason, the regression coefficients are not susceptible to complaints of "over-fitting." In essence, they use up just two degrees of freedom from a data set with more than 7,000 data points.
18. The histogram and its Gaussian fit also contain the "reference" subject's sensitivity, equal to +0.00.
19. For this observed distribution of sensitivities, the Shapiro-Wilk's W statistic equals 0.97, with a p-value of 0.50. That p-value shows no cause to reject that the observed distribution is Gaussian, and so it was accepted as Gaussian.
20. Note that duration of residence would be difficult to determine during application of these results at an airport under study. In contrast, this variable might truly contribute to each subject's aircraft-noise sensitivity. By dropping it, residence duration is thereby subsumed into subject awakening sensitivity, which can be taken into account (through the histogram of Fig. 5) during airport application of these results.
21. M. Brink et al., "Effects of early morning aircraft overflights on sleep and implications for policy making," EuroNoise 2006, Tampere, Finland (2006).
22. This equation is strictly true when the probabilities of awakening are independent, from one aircraft to the next. See discussion in Sec. 6.
23. In the same studies whose findings are re-analyzed here, aircraft-induced awakenings occurred at approximately half the rate of spontaneous (non-aircraft) awakenings. Although not founded in research, the authors find it reasonable that these more-numerous spontaneous awakenings are fundamentally less disruptive to the sleeper, since spontaneous awakenings are "volunteered," rather than "forced."
24. The authors do not contend that "awakening at least once per night" is an all-encompassing sleep-disturbance metric - only that it is one possible metric that is easily understood by airport neighbors.
25. It is pure happenstance that the number of population groups (33) equals the number of sleep subjects in the "regression" data.
26. J. B. Ollerhead et al., "Report of a field study of aircraft noise and sleep disturbance," Department of Transport, Civil Aviation Authority, U.K. (1992). Arousal was determined as "any EEG-determined arousal to wakefulness" of 15 seconds or more, or movement time lasting 10 seconds or more.
27. W. Passchier-Vermeer et al., "Sleep disturbance and aircraft noise exposure: Exposure-effect relationships," Report TNO 2002.027, Division Public Health, The Netherlands (2002).
28. M. Basner et al., "Effects of nocturnal aircraft noise - Volume 1 : Executive summary," Report 2004-07/E, German Aerospace Center (DLR), Institute of Aerospace Medicine, Flightphysiology Department, Cologne, Germany (2004).
29. In addition, re-scheduling aircraft to fly earlier in the night is likely beneficial to the "recovery function" of sleep. In particular, impeded slow-wave sleep in the first part of the night can more easily be made up in the second part, rather than the other way round. Moreover, when awakened in the early morning, people will have more trouble falling asleep again, so will more likely remember their awakening and potentially convert that memory to "annoyance."
30. Regression 1 numerically equals Regression 3 for the average indoor SEL (72 dB), the average event timing (slightly under three hours), and the average sensitivity (50-percentile person).
31. Each person's dose-response curve has a particular left-right shift, depending upon that persons' sleep stage at the moment, emotional state, sub-conscious interpretation of the potential meaning of the noise, total time in bed, fatigue level, use of sleeping medication, and a host of other factors that pertain to that person, at that moment. All these matters affect the horizontal location of that person's dose-response curve at the moment - that is, that person's sensitivity to noise-induced awakening.
32. Note that the sleep-study data were obtained in neighborhoods with "stable" aircraft exposure. For this reason, if aircraft noise is suddenly decreased by some action of the airport, the short-term reduction in awakening may not mirror the slope of this dose-response curve. Transitional instabilities may intervene. We expect, however, that after the neighborhood returns to aircraft-noise stability (perhaps several years?), that the SEL slope will approximate the accrued long-term reduction in aircraft awakenings. In addition, remember that the dose-response curve does not predict individual reactions, person by person, but only the distribution of reactions over large totals of people.
33. Note that the resulting 83% is not what Eqn. (5) would predict. Instead, it is the simple average of the values from the top and bottom curves, which represent a very simplified awakening-sensitivity distribution in this explanatory figure.
34. One reviewer stated that the sampling of the original sleep studies was not random, but was "purposive," instead. In response, purposive samples are often used to represent larger populations. Because the "validation" data so clearly validated results from the "regression data," we contend that the purposive nature of subject selection did not introduce significant systematic subject bias. To the extent that it did, then this analysis - as well as the analysis of the original sleep studies - may not be representative. In addition, this study used what data were available to the authors. Until the methods of this paper can be applied to additional data, the authors suggest that there is considerable value in not only the method but in the findings.
35. A. Krupnik et al., "Not a sure thing: Making regulatory choices under uncertainty," Resources for the Future, Inc., www.rff.org(under Contract 68-W-02-046 to the U.S. Environmental Protection Agency), Washington DC (2006).
36. Office of Management and Budget (OMB), "Regulatory analysis as required under Section 6(a)(3)(c) of Executive Order12866, 'Regulatory Planning and Review,' the Regulatory Right-to-Know Act, and a variety of related authorities," Circular A-4, Washington DC (2003).
37. M. Basner, "Markov state transition models for the prediction of changes in sleep structure induced by aircraft noise," Research report 2006-07, German center for air and space travel (Deutsches Zentrum für Luft- und Raumfahrt, DLR), Institute of Aerospace Medicine, Cologne. Available at www.eid.dlr.de/me/Institut/Abteilungen/Flugphysiologie/Fluglaerm/ Lit_Laermw_Reports/FB_2006-07_markov.pdf.
38. M. Brink et al., "Bestimmung lärminduzierter Aufwachwahrscheinlichkeiten in der Nachtlärmwirkungsforschung und Anwendung entsprechender Wirkungsmodelle für Prognosezwecke," Report to the Workshop 'Aufwachwahrscheinlichkeit,' Swiss Federal Institute of Technology Zurich, 21-22 April 2006 (English translation in preparation). Available at http://e-collection.ethbib.ethz.ch/ecol-pool/bericht/bericht_485. pdf.
39. International Standards Organization (ISO), "Guide to the expression of uncertainty in measurement," Section 5.2 for correlated input (corrected and reprinted 1995).
40. I. S. Guttman, W. Wilks and J. S. Hunter, Introductory engineering statistics: Third edition, Section 9.8, John Wiley & Sons (1982).
41. y in that region, as shown. This definition of n allows it to compensate for binary over-dispersion - a common occurrence.
42. 2 values are not relevant single-number assessments of logistic regressions. Relevant single-number assessments - such as the regression's logliklihood and Akaeka Information Criterion - are not reported here. Instead, the more-informative uncertainty bounds of Figs. 7 and 7 serve to judge the quality of the regression, as a function of specific input values of the independent parameters. As those figures show, regression uncertainty varies greatly over different ranges of those parameters.