A multivariate regression model based on an optimal partition of predictors (MRBOP) is presented. The proposed model aims at identifying a few groups of correlated explanatory variables which similarly predict the responses. In MRBOP, linear combinations of the variables in the same groups define latent predictors easy to be interpreted representing underlying features of the data with high prediction power. The model is formalized in a least squares framework optimizing a penalized quadratic objective function. An alternating least-squares (ALS) algorithm for fitting the MRBOPmodel is proposed, and the performance of the new methodology is evaluated in simulation studies.
Multivariate regression model based on a partition of predictors / Martella, Francesca; Vicari, Donatella; Vichi, Maurizio. - STAMPA. - (2011). (Intervento presentato al convegno 7th Conference on Statistical Computation and Complex Systems (SCO) tenutosi a Padova, Italy nel 19-21 settembre, 2011).
Multivariate regression model based on a partition of predictors.
MARTELLA, Francesca;VICARI, Donatella;VICHI, Maurizio
2011
Abstract
A multivariate regression model based on an optimal partition of predictors (MRBOP) is presented. The proposed model aims at identifying a few groups of correlated explanatory variables which similarly predict the responses. In MRBOP, linear combinations of the variables in the same groups define latent predictors easy to be interpreted representing underlying features of the data with high prediction power. The model is formalized in a least squares framework optimizing a penalized quadratic objective function. An alternating least-squares (ALS) algorithm for fitting the MRBOPmodel is proposed, and the performance of the new methodology is evaluated in simulation studies.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
