Statistics Seminar - Finding the number of clusters with multivariate time series

Título: Encontrar el número de clusters con series temporales multivariantes
Expositor: Daniel Peña, profesor de la Universidad Carlos III de Madrid y es invitado del Instituto de Cálculo y la Maestría en Estadística Matemática.

Día: Miércoles 22/2
Hora: 17:30 hs
Lugar: pab 0 - 0  Aula 1101

ABSTRACT
Clustering scalar time series using their univariate properties and a hierarchical method is considered.  Two major issues, in this case, are to detect the existence of multiple clusters and to determine their number if exist. In this paper, we propose a  new test statistic
for detecting the existence of multiple clusters in a time-series data set and a new procedure to determine its number when they exist.
The proposed method is based on the jumps, i.e.,  the increments, in the heights of the dendrogram when a hierarchical clustering is applied to the data. We use parametric bootstraps to obtain a reference distribution of the test statistics and propose an iterative procedure to find the number of clusters. The clusters found are internally homogeneous according to the test statistics used in the analysis. The performance of the proposed procedure in finite samples is investigated by Monte Carlo simulations and illustrated by some empirical examples. Comparisons with some existing methods for selecting the number of clusters are also investigated.