Transposable regularized covariance models with an application to missing data imputation
From MaRDI portal
Publication:993250
DOI10.1214/09-AOAS314zbMATH Open1194.62079arXiv0906.3465WikidataQ27347321 ScholiaQ27347321MaRDI QIDQ993250FDOQ993250
Authors: Genevera I. Allen, Robert Tibshirani 
Publication date: 10 September 2010
Published in: The Annals of Applied Statistics (Search for Journal in Brave)
Abstract: Missing data estimation is an important challenge with high-dimensional data arranged in the form of a matrix. Typically this data matrix is transposable, meaning that either the rows, columns or both can be treated as features. To model transposable data, we present a modification of the matrix-variate normal, the mean-restricted matrix-variate normal, in which the rows and columns each have a separate mean vector and covariance matrix. By placing additive penalties on the inverse covariance matrices of the rows and columns, these so-called transposable regularized covariance models allow for maximum likelihood estimation of the mean and nonsingular covariance matrices. Using these models, we formulate EM-type algorithms for missing data imputation in both the multivariate and transposable frameworks. We present theoretical results exploiting the structure of our transposable models that allow these models and imputation methods to be applied to high-dimensional data. Simulations and results on microarray data and the Netflix data show that these imputation techniques often outperform existing methods and offer a greater degree of flexibility.
Full work available at URL: https://arxiv.org/abs/0906.3465
Recommendations
- Missing values: sparse inverse covariance estimation and an extension to sparse regression
- An Imputation–Regularized Optimization Algorithm for High Dimensional Missing Data Problems and Beyond
- High-dimensional covariance matrix estimation with missing observations
- Pattern alternating maximization algorithm for missing data in high-dimensional problems
- Minimax rate-optimal estimation of high-dimensional covariance matrices with incomplete data
Cites Work
- Covariance-regularized regression and classification for high dimensional problems
- The mle algorithm for the matrix normal distribution
- Variable Selection via Nonconcave Penalized Likelihood and its Oracle Properties
- Maximum likelihood estimation via the ECM algorithm: A general framework
- Title not available (Why is that?)
- Choosing starting values for the EM algorithm for getting the highest likelihood in multivariate Gaussian mixture models
- Sparse inverse covariance estimation with the graphical lasso
- Sparse permutation invariant covariance estimation
- Exact matrix completion via convex optimization
- Multiple Imputation After 18+ Years
- Title not available (Why is that?)
- Component selection and smoothing in multivariate nonparametric regression
- Title not available (Why is that?)
- Stochastic versions of the em algorithm: an experimental study in the mixture case
- Correlated \(z\)-values and the accuracy of large-scale statistical estimates
- Transposable regularized covariance models with an application to missing data imputation
- Are a set of microarrays independent of each other?
Cited In (41)
- PLS for Big Data: a unified parallel algorithm for regularised group PLS
- Hypothesis Testing of Matrix Graph Model with Application to Brain Connectivity Analysis
- Existence and uniqueness of the maximum likelihood estimator for models with a Kronecker product covariance structure
- Scalable Bayesian matrix normal graphical models for brain functional networks
- Transposable regularized covariance models with an application to missing data imputation
- A generalized least-square matrix decomposition
- The mixed Lipschitz space and its dual for tree metrics
- Consistency of large dimensional sample covariance matrix under weak dependence
- Pattern alternating maximization algorithm for missing data in high-dimensional problems
- Covariance estimation via sparse Kronecker structures
- Co-clustering of spatially resolved transcriptomic data
- Gaussian and robust Kronecker product covariance estimation: existence and uniqueness
- Gemini: graph estimation with matrix variate normal instances
- Model selection and estimation in the matrix normal graphical model
- Kronecker-structured covariance models for multiway data
- Separable factor analysis with applications to mortality data
- Concentration of measure bounds for matrix-variate data with missing values
- Estimating high-dimensional covariance and precision matrices under general missing dependence
- Inferring Phenotypic Trait Evolution on Large Trees With Many Incomplete Measurements
- Title not available (Why is that?)
- Imputation and low-rank estimation with missing not at random data
- An expectation-maximization algorithm for the matrix normal distribution with an application in remote sensing
- Mixed Hölder matrix discovery via wavelet shrinkage and Calderón-Zygmund decompositions
- Many-sample tests for the equality and the proportionality hypotheses between large covariance matrices
- Maximum likelihood estimation for matrix normal models via quiver representations
- Unifying and generalizing methods for removing unwanted variation based on negative controls
- Testing high-dimensional mean vector with applications. A normal reference approach
- Testing the mean matrix in high-dimensional transposable data
- A constrained matrix-variate Gaussian process for transposable data
- Missing values: sparse inverse covariance estimation and an extension to sparse regression
- Matrix Completion, Counterfactuals, and Factor Analysis of Missing Data
- Sparse Matrix Graphical Models
- Mixture of multivariate Gaussian processes for classification of irregularly sampled satellite image time-series
- Existence and uniqueness of the Kronecker covariance MLE
- Graphical model selection and estimation for high dimensional tensor data
- Recovering networks from distance data
- Autoregressive identification of Kronecker graphical models
- Hypothesis testing for the covariance matrix in high-dimensional transposable data with Kronecker product dependence structure
- A Penalized Likelihood Method for Classification With Matrix-Valued Predictors
- Detecting column dependence when rows are correlated and estimating the strength of the row correlation
- Sampling, denoising and compression of matrices by coherent matrix organization
Uses Software
This page was built for publication: Transposable regularized covariance models with an application to missing data imputation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q993250)
