Semi-Independent Sliced Inverse Regression

Kevin Li (1)

1 BJTU - Beijing Jiaotong University

This paper deals with dimension reduction in regression for large dataset. A new method based on the sliced inverse regression approach is introduced, called semi-independent regularized sliced inverse regression. Our method can handle highly correlated variables. Asymptotic properties are established under some linearity conditions. An application on an economic dataset shows that our method outperformes the non-stationary factor model.

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Kevin Li. Semi-Independent Sliced Inverse Regression. 2014. ⟨hal-00960343⟩