A general structure for the problem of selection of variables in regression is proposed using the decision theory framework. In particular, some results for the choice of the best linear normal homocedastic model are obtained when the main purpose is either to specify the predictive distribution over the response variable or to obtain a point estimate of it. A comparison of our results with the most widespread classical ones is presented.
@article{urn:eudml:doc:40413,
title = {Elecci\'on de variables en regresi\'on lineal: un problema de decisi\'on.},
journal = {Trabajos de Estad\'\i stica},
volume = {1},
year = {1986},
pages = {3-16},
language = {es},
url = {http://dml.mathdoc.fr/item/urn:eudml:doc:40413}
}
Bermúdez, José D. Elección de variables en regresión lineal: un problema de decisión.. Trabajos de Estadística, Tome 1 (1986) pp. 3-16. http://gdmltest.u-ga.fr/item/urn:eudml:doc:40413/