E. García-Portugués, R. Crujeiras Casais, W. González-Manteiga

Directional statistics deal with data lying on the $q$-dimensional sphere $\Omega_q=\{\mathbf{x}\in\mathbb{R}^{q+1}:|\mathbf{x}|=1\}$, with the most common cases corresponding to the circle ($q=1$) and the sphere ($q=2$). In many applied fields, it may be interesting to assess the relation between directional and linear random variables. For instance, in environmental protection, pollutants concentration and wind direction joint structure may be used to detect emission sources. In this work, an estimator for the joint density of a directional-linear random variable $(\mathbf{X},Z)$ with support in $\Omega_q\times\mathbb{R}$ is proposed. This estimator is based on a directional-linear kernel product and expressions for bias, variance and MSE are derived. Optimal smoothing parameters in terms of the AMISE criterion is also provided. The finite sample properties of the estimator are explored throughout a simulation study. Finally, the estimator is illustrated with real data examples.

Palabras clave: directional-linear, circular-linear, directional data, circular data, kernel density estimation

Fichero de la comunicación:

Programado

VA7 Estadística no paramétrica
20 de abril de 2012  09:00
Sala Roma II

Análisis de similaridad de k muestras

P. C. Alvarez Esteban

Bandwidth selection for mean shift clustering

J. E. Chacón

Métodos estadísticos para el análisis 3D de moléculas en forma de anillo

M. Kessler, J. Pérez Pérez