Robust estimation of scale for local linear temporal trends

Rok, strany: 2003, 87 - 101

Online monitoring data measured in intensive care exhibit periods of relative constancy corresponding to a steady state and interruptions representing slow monotonic trends and level shifts. These signals are overlaid with a high level of noise and many measurement artifacts. Methods are needed that allow a fast and reliable denoising of the data and separate artifacts from information on clinical relevant changes in the patient's condition. The underlying signal can be approximated by fitting a local linear trend within a moving time window using, e.g., the repeated median [P. L. Davies, R. Fried and U. Gather: Robust signal extraction from on-line monitoring data, Technical Report 2/2002, SFB 475, University of Dortmund, Germany (2002)]. We compare the finite-sample performance of robust functionals for scale estimation in trend periods as these can be used to detect outliers and level shifts.

ISO 690:

Gather, U., Fried, R. 2003. Robust estimation of scale for local linear temporal trends. In Tatra Mountains Mathematical Publications, vol. 26, no.1, pp. 87-101. 1210-3195.

APA:

Gather, U., Fried, R. (2003). Robust estimation of scale for local linear temporal trends. Tatra Mountains Mathematical Publications, 26(1), 87-101. 1210-3195.