This Working Paper presents some preliminary results for a new goodness-of- t method for VARMA(p,q) models. Relations between least squares residuals and true errors are re-examined, and a new family of statistics is proposed. A new goodness-of- t process is also suggested, that can be seen as...

This paper explores some properties of the quadratic subspace, a tool for dimension reduction in discriminant analysis (Velilla, 2008, 2010). This linear manifold has a fairly complex structure, and it may sometimes include components with both mean and covariance separation properties. In this...

En este trabajo se formalizan algunos resultados relativos al comportamiento asintótico del correlograma residual que, en la literatura de series temporales, aparecen recurrentemente y son de gran utilidad para la obtención y demostración de algunas propuestas de interés. Como ejemplo, se...

Strong consistency of linear discriminant analysis is established under wide assumptions on the class conditional densities. Robustness to the presence of a mild degree of class dispersion heterogeneity is also analyzed. Results obtained may help to explain analytically the frequent good...

The concept of quadratic subspace is introduced as a helpful tool for dimension reduction in quadratic discriminant analysis (QDA). It is argued that an adequate representation of the quadratic subspace may lead to better methods for both data representation and classification. Several...

A dimension reduction method in kernel discriminant analysis is presented, based on the concept of dimension reduction subspace. Examples of application are discussed.

Dimension reduction transformations in discriminant analysis are introduced. Their properties, as well as sufficient conditions for their characterization, are studied. Special attention is given to the continuous case, of particular importance in applications. An effective data based dimension...

In this paper, a new procedure for testing the number of linear components in a general regression problem is introduced. It is based on a nonparametric estimate of the covariance matrix of the inverse regression curve. A review of previous dimension tests is also presented.

A first-order bootstrap method for assessing the dimension of a general regression problem is presented. The proposed approach combines both formal and graphical inference procedures.

The limit behavior of the conditional probability of error of linear and quadratic discriminant analyses is studied under wide assumptions on the class conditional distributions. Results obtained may help to explain analytically the behavior in applications of linear and quadratic discrimination...