A Bootstrap Method for the Analysis of Physiological Data in Uncontrolled Settings

Abstract

There are difficulties associated with analyzing physiological data sets, especially those that are collected at high frequency and those obtained from uncontrolled, non-laboratory task settings. Difficulties include the fact that the data are not independent and not always normally distributed. The authors used a bootstrap ANOVA method to analyze physiological responses collected during the performance of field science deployments in a Mars exploration analog study, and to identify differences in the physiological responses by different task performers in different tasks. The method used addressed the dependence and unknown a priori distributions of the data and provided information as to whether or not the collected responses to each task were statistically significant. Despite the fact that the independence and distribution difficulties were addressed, there were still limitations to the method, which include: the reduction of the power of the analysis, the increased computational resources required, and the limitation that the method does not output p-values to support traditional human factors analyses of whether or not statistical differences exist.