Authors
Slaven-JE; Andrew-ME; Violanti-JM; Burchfiel-CM; Vila-BJ
Source
Physiol Meas 2006 Apr; 27(4):413-423
Abstract
Accelerometer data quality can be inadequate due to data corruption or to non-compliance of the subject with regard to study protocols. We propose a simple statistical test to determine if accelerometer data are of good quality and can be used for analysis or if the data are of poor quality and should be discarded. We tested several data evaluation methods using a group of 105 subjects who wore Motionlogger actigraphs (Ambulatory Monitoring, Inc.) over a 15 day period to assess sleep quality in a study of health outcomes associated with stress among police officers. Using leave-one-out cross-validation and calibration-testing methods of discrimination statistics, error rates for the methods ranged from 0.0167 to 0.4046. We found that the best method was to use the overall average distance between consecutive time points and the overall average mean amplitude of consecutive time points. These values gave us a classification error rate of 0.0167. The average distance between points is a measure of smoothness in the data, and the average mean amplitude between points gave an average reading. Both of these values were then normed to determine a final statistic, K, which was then compared to a cut-off value, K(C), to determine data quality.
Keywords
Statistical-analysis; Sleep-disorders; Sleep-deprivation; Occupational-health; Stress; Police-officers; Physiological-measurements; Physiological-testing
Contact
NIOSH, MS 4050, 1095, Willowdale Rd, Morgantown, WV 26501, USA
Document Type
Journal Article
Email Address
cto8@cdc.gov
Source Name
Physiological Measurement