Distribution of subjective assessments in a controlled aircraft environment

摘要

Carl von Ossietzky University of Oldenburg, Institute of Physics, Acoustics group, Carl von Ossietzky Str. 9-11, D-26129 Oldenburg, Germany;Medical University Vienna, Unit for Psychology of Ergonomics, Environment, Health and Performance at the Institute for Environmental Hygiene, Centre for Public Health,Kinderspitalgasse 15, A-1095 Vienna, Austria;%Within the project ICE ("Ideal Cabin Environment", funded by the European Commission under contract no. AST4-CT-2005-516131), the impact of different environmental conditions on comfort was investigated by 1500 participants in total. The ICE consortium was made up of 15 members from eight countries. Simulated flights of 8 hours each with 1100 passengers on 29 days were conducted in a large-scale aircraft cabin environment facility near Munich, Germany in winter 2006/2007. During these tests, typical in-flight conditions for noise, vibration, temperature, humidity, and pressure were presented with different level combinations on different days in order to analyze possible effects of these parameters on participants' comfort and well-being. The environmental conditions and passengers' physiological response were monitored by sensors, while psychological response was recorded using PDA-based questionnaires. For our share of evaluation, extensive statistical methods were used, applied on a flight-by-flight analysis. As one outcome, the possibility to describe a large class of responses by a function based on the Poisson distribution is shown. An approximation of observed data was performed using the d_(max) test criteria of the Kolmogorov-Smirnov goodness-of-fit test. In addition, R as the variance-to-mean ratio (VMR) was determined and compared to the quality of the predictions with a significant equality to the observed data. For subjective ratings on a unipolar scale and a calculated VMR of R ≥ 0.72 in raw distribution data, a success rate of nearly 95% in representing the observed data distribution with one single parameter was identified. This approach allows for a representation of an average participant's impression (answer distributions of questionnaire data) of a defined environmental condition and enables for a prediction of not measured, interim situations.