Resumen de New insights into evaluation of regression models through a decomposition of the prediction errors: application to near-infrared spectral data
María Isabel Sánchez Rodríguez, Elena M. Sánchez López, José María Caridad y Ocerín, Alberto Marinas Aramendia, José María Marinas Rubio, Francisco José Urbano Navarro
This paper analyzes the performance of linear regression mo dels taking into account usual criteria such as the number of principal components or latent factors , the goodness of fit or the predictive capability. Other comparison criteria, more common in an ec onomic context, are also considered:
the degree of multicollinearity and a decomposition of the m ean squared error of the prediction which determines the nature, systematic or random, of the pr ediction errors. The applications use real data of extra-virgin oil obtained by near-infrared spe ctroscopy. The high dimensionality of the data is reduced by applying principal component analysi s and partial least squares analysis.
A possible improvement of these methods by using cluster ana lysis or the information of the relative maxima of the spectrum is investigated. Finally, o btained results are generalized via cross- validation and bootstrapping.
