Predicting the vulnerability to sleep deprivation

Abstract

We all know that sleep deprivation can adversely affect the brain and cognitive function. We are slower at responding, making more mistakes, etc. Yet, some people are more vulnerability to sleep deprivation than others. The project aim is to answer such difference by predicting the vulnerability of a person with sleep deprivation from the reaction times collected from Psychomotor Vigilance Task which is a sustained-attention reaction time task that measure the time that subjects response to a visual stimulus. For each set of RT collected from a subject, there is a combination of four parameters that parameterize the data. The four parameters are mean and standard deviation of decision time and non-decision time of the subject that generates the RT set. An estimator has been developed to estimate those parameters from a RT data set using mean square error technique. This method is to try to minimize the difference of the probability density function(pdf) of observed RT and the pdf of simulated RT, which was simulated from the initial guess and adjusted parameters during estimating process. An evaluation metric for the estimator which is the Cramer-Rao Lower bound is developed to evaluate the accuracy of the estimator. Cramer-Rao Lower bound is the minimum variance of an unbiased estimator. The efficient of the estimator which could tell how close of the estimator could be approach to minimum variance can be figured out. With the reliable evaluation metrics, the estimator would be evaluated accurately and hence would be more reliable.