TY - JOUR
T1 - Projective resampling estimation of informative predictor subspace for multivariate regression
AU - Ko, Sojin
AU - Yoo, Jae Keun
N1 - Funding Information:
For Jae Keun Yoo, this work was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Korean Ministry of Education (NRF-2021R1F1A1059844). The authors are grateful to the two referees and the Associate Editor for their many helpful comments and suggestions.
Publisher Copyright:
© 2022, Korean Statistical Society.
PY - 2022/12
Y1 - 2022/12
N2 - In this paper, a paradigm to estimate the so-called informative predictor subspace (Yoo in Statistics 50:1086–1099, 2016) for multivariate regression is proposed. For this, as a primary target subspace, a projective resampling informative predictor subspace is newly developed. The projective resampling informative predictor subspace is constructed based on a projection resampling method by Li et al. (2008), and it has advantage that it is smaller than the original informative predictor subspace but contains the central subspace. To estimate the new target subspace, the three approaches of projective resampling, coordinate, and coordinate-projective resampling mean methods are proposed. The three methods are investigated via various numerical studies, which confirm their potential usefulness in practice.
AB - In this paper, a paradigm to estimate the so-called informative predictor subspace (Yoo in Statistics 50:1086–1099, 2016) for multivariate regression is proposed. For this, as a primary target subspace, a projective resampling informative predictor subspace is newly developed. The projective resampling informative predictor subspace is constructed based on a projection resampling method by Li et al. (2008), and it has advantage that it is smaller than the original informative predictor subspace but contains the central subspace. To estimate the new target subspace, the three approaches of projective resampling, coordinate, and coordinate-projective resampling mean methods are proposed. The three methods are investigated via various numerical studies, which confirm their potential usefulness in practice.
KW - Clustering mean method
KW - Fused estimation
KW - Informative predictor subspace
KW - K-means clustering
KW - Sufficient dimension reduction
UR - http://www.scopus.com/inward/record.url?scp=85134507070&partnerID=8YFLogxK
U2 - 10.1007/s42952-022-00178-0
DO - 10.1007/s42952-022-00178-0
M3 - Article
AN - SCOPUS:85134507070
SN - 1226-3192
VL - 51
SP - 1117
EP - 1131
JO - Journal of the Korean Statistical Society
JF - Journal of the Korean Statistical Society
IS - 4
ER -