A weighted distance for short time series

Time-varying data have often been analyzed by methods that do not explicitly take into account time related changes. To overcome this problem, we segmented a time series in many components, each of which yields a separate unit of information that can act on a stand-alone basis for the computation of the distance function. As a consequence, the comparison of two time series becomes a comparison of two sets of matrices. To simplify the comparison, the procedure Distatis can be very helpful because it yields a compromise distance matrix that can be analyzed by using the usual techniques of cluster analysis.