scientific article; zbMATH DE number 1047867

From MaRDI portal
Publication:4347946





A simple linear regression distribution is treated here as a distribution for one cluster. The term cluster is used here in a non-traditional sense. This term does not depend immediately on the distance between the points, which create the clusters. The problems can be also described using the term ``outliers (an outliers region). Linear regression distributions with different parameters are supposed to be adequate for different parts of the dataset (for different clusters). There are two different formulations of models of cluster analysis: mixture models and the fixed point clusters. The new method -- fixed point cluster analysis -- enables the analysis of data with an unknown number of clusters and outliers. Both types of models can be formulated using fixed regressors or stochastic regressors. Every fixed point cluster corresponds to parameters, which can be interpreted as an estimation of certain regression parameters. The special case of the fixed point cluster method is the change point problem. The identifiability of the model parameters plays an important role for consistent estimation in cluster models. The identifiability is also studied for the considered models. The general problem of the adequacy of models is also discussed. The numerical calculations are described and discussed in detail. The simulated behaviour of various methods is studied in various data situations.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4347946)