Dimension Reduction for Time Series from Intensive Care
Ursula Gather, Vivian Lanius · RePEc: Research Papers in Economics · 2003
A modified version of principal component analysis (PCA) for time series is investigated. The approach is in the frequency domains as in Brillinger (1975). Available knowledge on the subject matter can be incorporated via rotational methods. This eases the interpretation of the obtained component series. An application to the hemodynamic data from intensive care yields clinically meaningful component series of low dimension. We describe the results from this application and compare them with those obtained standard PCA.