Using an Informative Missing Data Model to Predict the Ability to Assess Recovery of Balance Control after SpaceflightAstronauts show degraded balance control immediately after spaceflight. To assess this change, astronauts' ability to maintain a fixed stance under several challenging stimuli on a movable platform is quantified by "equilibrium" scores (EQs) on a scale of 0 to 100, where 100 represents perfect control (sway angle of 0) and 0 represents data loss where no sway angle is observed because the subject has to be restrained from falling. By comparing post- to pre-flight EQs for actual astronauts vs. controls, we built a classifier for deciding when an astronaut has recovered. Future diagnostic performance depends both on the sampling distribution of the classifier as well as the distribution of its input data. Taking this into consideration, we constructed a predictive ROC by simulation after modeling P(EQ = 0) in terms of a latent EQ-like beta-distributed random variable with random effects.
Document ID
20080010633
Acquisition Source
Johnson Space Center
Document Type
Conference Paper
Authors
Feiveson, Alan H. (NASA Johnson Space Center Houston, TX, United States)
Wood, Scott J. (Universities Space Research Association Houston, TX, United States)
Jain, Varsha (Imperial Coll. of London London, United Kingdom)