Partial least squares for dependent data

被引:14
作者
Singer, Marco [1 ]
Krivobokova, Tatyana [1 ]
Munk, Axel [1 ]
De Groot, Bert [2 ]
机构
[1] Univ Gottingen, Inst Math Stochast, Goldschmidtstr 7, D-37077 Gottingen, Germany
[2] Max Planck Inst Biophys Chem, Fassberg 11, D-37077 Gottingen, Germany
关键词
Dependent data; Latent variable model; Nonstationary process; Partial least squares; Protein dynamics; REGRESSION; PROTEINS; PLS; CHEMOMETRICS; PREDICTION; MATRICES; MOTIONS; MODELS;
D O I
10.1093/biomet/asw010
中图分类号
Q [生物科学];
学科分类号
07 ; 0710 ; 09 ;
摘要
We consider the partial least squares algorithm for dependent data and study the consequences of ignoring the dependence both theoretically and numerically. Ignoring nonstationary dependence structures can lead to inconsistent estimation, but a simple modification yields consistent estimation. A protein dynamics example illustrates the superior predictive power of the proposed method.
引用
收藏
页码:351 / 362
页数:12
相关论文
共 32 条
[1]   Time series analysis of collective motions in proteins [J].
Alakent, B ;
Doruker, P ;
Çamurdan, MC .
JOURNAL OF CHEMICAL PHYSICS, 2004, 120 (02) :1072-1088
[2]   A simulation study on comparison of prediction methods when only a few components are relevant [J].
Almoy, T .
COMPUTATIONAL STATISTICS & DATA ANALYSIS, 1996, 21 (01) :87-107
[3]  
[Anonymous], 1966, Research Papers in Statistics: Festschrift for J. Neyman
[4]  
[Anonymous], 2010, ADV NEURAL INFORM PR
[5]  
[Anonymous], JMLR WORKSHOP C P
[6]   Regularized estimation of large covariance matrices [J].
Bickel, Peter J. ;
Levina, Elizaveta .
ANNALS OF STATISTICS, 2008, 36 (01) :199-227
[7]  
Brockwell P., 1998, Time Series: Theory and Methods
[8]  
de Groot BL, 1998, PROTEINS, V31, P116, DOI 10.1002/(SICI)1097-0134(19980501)31:2<116::AID-PROT2>3.0.CO
[9]  
2-K
[10]   METHODOLOGY AND THEORY FOR PARTIAL LEAST SQUARES APPLIED TO FUNCTIONAL DATA [J].
Delaigle, Aurore ;
Hall, Peter .
ANNALS OF STATISTICS, 2012, 40 (01) :322-352