Sparse principal component analysis with missing observations

DOI10.1007/978-3-0348-0490-5_20MaRDI QIDQ2840346zbMATH OpenOpenAlexFDO

Authors K. Lounici

Publication date 18 July 2013

Published in Progress in Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1205.7060

zbMATH Keywords

oracle inequalitiesinformation-theoretic lower boundslow-rank covariance matrix

Mathematics Subject Classification ID

Statistical aspects of information-theoretic topics (62B10) Factor analysis and principal components; correspondence analysis (62H25)

Abstract: In this paper, we study the problem of sparse Principal Component Analysis (PCA) in the high-dimensional setting with missing observations. Our goal is to estimate the first principal component when we only have access to partial observations. Existing estimation techniques are usually derived for fully observed data sets and require a prior knowledge of the sparsity of the first principal component in order to achieve good statistical guarantees. Our contributions is threefold. First, we establish the first information-theoretic lower bound for the sparse PCA problem with missing observations. Second, we propose a simple procedure that does not require any prior knowledge on the sparsity of the unknown first principal component or any imputation of the missing observations, adapts to the unknown sparsity of the first principal component and achieves the optimal rate of estimation up to a logarithmic factor. Third, if the covariance matrix of interest admits a sparse first principal component and is in addition approximately low-rank, then we can derive a completely data-driven procedure computationally tractable in high-dimension, adaptive to the unknown sparsity of the first principal component and statistically optimal (up to a logarithmic factor).

Recommendations

• Sparse principal component analysis with missing observations
• Sparse PCA: optimal rates and adaptive estimation
• Sparse principal component analysis and iterative thresholding
• Practical approaches to principal component analysis in the presence of missing values
• Minimax sparse principal subspace estimation in high dimensions

Cited in

(15)

• Discussion of ``Estimating structured high-dimensional covariance and precision matrices: optimal rates and adaptive estimation
• Minimax sparse principal subspace estimation in high dimensions
• Dynamic principal component analysis with missing values
• Sparse principal component analysis with missing observations
• Optimal estimation and rank detection for sparse spiked covariance matrices
• Sparse PCA: optimal rates and adaptive estimation
• Minimax rate-optimal estimation of high-dimensional covariance matrices with incomplete data
• Subspace estimation from unbalanced and incomplete data matrices: \({\ell_{2,\infty}}\) statistical guarantees
• New asymptotic results in principal component analysis
• scientific article; zbMATH DE number 3911503 (Why is no real title available?)
• Recovering PCA and sparse PCA via hybrid-\((\ell_1,\ell_2)\) sparse sampling of data elements
• ECA: High-Dimensional Elliptical Component Analysis in Non-Gaussian Distributions
• Rejoinder of ``Estimating structured high-dimensional covariance and precision matrices: optimal rates and adaptive estimation
• Sparsistency and agnostic inference in sparse PCA
• Inference for heteroskedastic PCA with missing data

This page was built for publication: Sparse principal component analysis with missing observations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2840346)