Abstract:
|
A variety of diagnostic devices often require a need for detection and localization of multiple events within a subject over a time course used mainly for monitoring of events e.g. the detection of events like seizures in temporal intervals from a scalp EEG reading. It is possible to characterize event detection in the presence of a reference standard and a priori time window overlap or proximity criteria, by sensitivity which is the probability of detecting the event given that the event has actually occurred. However, specificity, which is the probability of a non-event given that the event did not occur, cannot be easily estimated. Instead, counting the false positives and reporting the rate of false positives is an alternative to evaluate device performance. A method to assess that the device is not randomly marking events is presented along with discussions of sample size calculations that address the intra-cluster variability due to multiple events per subject.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.