A convex pseudolikelihood framework for high dimensional partial correlation estimation with convergence guarantees
Abstract
Sparse high dimensional graphical model selection is a topic of much interest in modern day statistics. A popular approach is to apply l1-penalties to either parametric likelihoods, or regularized regression/pseudolikelihoods, with the latter having the distinct advantage that they do not explicitly assume Gaussianity. As none of the popular methods proposed for solving pseudolikelihood-based objective functions have provable convergence guarantees, it is not clear whether corresponding estimators exist or are even computable, or if they actually yield correct partial correlation graphs. We propose a new pseudolikelihood-based graphical model selection method that aims to overcome some of the shortcomings of current methods, but at the same time retain all their respective strengths. In particular, we introduce a novel framework that leads to a convex formulation of the partial covariance regression graph problem, resulting in an objective function comprised of quadratic forms. The objective is then optimized via a coordinate wise approach. The specific functional form of the objective function facilitates rigorous convergence analysis leading to convergence guarantees; an important property that cannot be established by using standard results, when the dimension is larger than the sample size, as is often the case in high dimensional applications. These convergence guarantees ensure thatmore »
- Authors:
-
- Univ. of Florida, Gainesville, FL (United States)
- Stanford Univ., CA (United States)
- Publication Date:
- Research Org.:
- Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)
- Sponsoring Org.:
- USDOE Office of Science (SC)
- OSTI Identifier:
- 1523947
- Grant/Contract Number:
- AC02-05CH11231; DMS‐1106084; DMS‐0906392; DMS‐CG 1025465; AGS‐1003823, DMS‐1106642, DMS CAREER‐1352656, NSA H98230‐11‐1‐0194; DARPA‐YFA N66001‐111‐4131; SMC‐DBNKY
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Journal of the Royal Statistical Society: Series B (Statistical Methodology)
- Additional Journal Information:
- Journal Name: Journal of the Royal Statistical Society: Series B (Statistical Methodology); Journal Volume: 77; Journal Issue: 4; Journal ID: ISSN 1369-7412
- Publisher:
- Royal Statistical Society - Wiley
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 97 MATHEMATICS AND COMPUTING; Sparse inverse covariance estimation; Graphical model selection; Soft thresholding; Partial correlation graph; Convergence guarantee; Generalized pseudo-likelihood; Gene regulatory network
Citation Formats
Khare, Kshitij, Oh, Sang-Yun, and Rajaratnam, Bala. A convex pseudolikelihood framework for high dimensional partial correlation estimation with convergence guarantees. United States: N. p., 2014.
Web. doi:10.1111/rssb.12088.
Khare, Kshitij, Oh, Sang-Yun, & Rajaratnam, Bala. A convex pseudolikelihood framework for high dimensional partial correlation estimation with convergence guarantees. United States. https://doi.org/10.1111/rssb.12088
Khare, Kshitij, Oh, Sang-Yun, and Rajaratnam, Bala. Fri .
"A convex pseudolikelihood framework for high dimensional partial correlation estimation with convergence guarantees". United States. https://doi.org/10.1111/rssb.12088. https://www.osti.gov/servlets/purl/1523947.
@article{osti_1523947,
title = {A convex pseudolikelihood framework for high dimensional partial correlation estimation with convergence guarantees},
author = {Khare, Kshitij and Oh, Sang-Yun and Rajaratnam, Bala},
abstractNote = {Sparse high dimensional graphical model selection is a topic of much interest in modern day statistics. A popular approach is to apply l1-penalties to either parametric likelihoods, or regularized regression/pseudolikelihoods, with the latter having the distinct advantage that they do not explicitly assume Gaussianity. As none of the popular methods proposed for solving pseudolikelihood-based objective functions have provable convergence guarantees, it is not clear whether corresponding estimators exist or are even computable, or if they actually yield correct partial correlation graphs. We propose a new pseudolikelihood-based graphical model selection method that aims to overcome some of the shortcomings of current methods, but at the same time retain all their respective strengths. In particular, we introduce a novel framework that leads to a convex formulation of the partial covariance regression graph problem, resulting in an objective function comprised of quadratic forms. The objective is then optimized via a coordinate wise approach. The specific functional form of the objective function facilitates rigorous convergence analysis leading to convergence guarantees; an important property that cannot be established by using standard results, when the dimension is larger than the sample size, as is often the case in high dimensional applications. These convergence guarantees ensure that estimators are well defined under very general conditions and are always computable. In addition, the approach yields estimators that have good large sample properties and also respect symmetry. Furthermore, application to simulated and real data, timing comparisons and numerical convergence is demonstrated. We also present a novel unifying framework that places all graphical pseudolikelihood methods as special cases of a more general formulation, leading to important insights.},
doi = {10.1111/rssb.12088},
journal = {Journal of the Royal Statistical Society: Series B (Statistical Methodology)},
number = 4,
volume = 77,
place = {United States},
year = {Fri Sep 26 00:00:00 EDT 2014},
month = {Fri Sep 26 00:00:00 EDT 2014}
}
Web of Science
Figures / Tables:
Works referenced in this record:
Gaussian Markov Distributions over Finite Graphs
journal, March 1986
- Speed, T. P.; Kiiveri, H. T.
- The Annals of Statistics, Vol. 14, Issue 1
Evidence for dynamically organized modularity in the yeast protein–protein interaction network
journal, June 2004
- Han, Jing-Dong J.; Bertin, Nicolas; Hao, Tong
- Nature, Vol. 430, Issue 6995
The Elements of Statistical Learning
book, January 2009
- Hastie, Trevor; Tibshirani, Robert; Friedman, Jerome
- Springer Series in Statistics
A well-conditioned estimator for large-dimensional covariance matrices
journal, February 2004
- Ledoit, Olivier; Wolf, Michael
- Journal of Multivariate Analysis, Vol. 88, Issue 2
Gene co-expression network topology provides a framework for molecular characterization of cellular state
journal, May 2004
- Carter, S. L.; Brechbuhler, C. M.; Griffin, M.
- Bioinformatics, Vol. 20, Issue 14
Interplay between BRCA1 and RHAMM Regulates Epithelial Apicobasal Polarization and May Influence Risk of Breast Cancer
journal, November 2011
- Maxwell, Christopher A.; Benítez, Javier; Gómez-Baldó, Laia
- PLoS Biology, Vol. 9, Issue 11
An Improved Iterative Proportional Scaling Procedure for Gaussian Graphical Models
journal, January 2011
- Xu, Ping-Feng; Guo, Jianhua; He, Xuming
- Journal of Computational and Graphical Statistics, Vol. 20, Issue 2
On Estimating the Expected Return on the Market: An Exploratory Investigation
preprint, February 1980
- Merton, Robert C.
- NBER Working Paper Series
From The Cover: Robustness, scalability, and integration of a wound-response gene expression signature in predicting breast cancer survival
journal, February 2005
- Chang, H. Y.; Nuyten, D. S. A.; Sneddon, J. B.
- Proceedings of the National Academy of Sciences, Vol. 102, Issue 10
The Structure and Function of Complex Networks
journal, January 2003
- Newman, M. E. J.
- SIAM Review, Vol. 45, Issue 2
Sparse inverse covariance estimation with the graphical lasso
journal, December 2007
- Friedman, J.; Hastie, T.; Tibshirani, R.
- Biostatistics, Vol. 9, Issue 3
On Portfolio Optimization: Forecasting Covariances and Choosing the Risk Model
preprint, March 1999
- Chan, Louis K. C.; Karceski, Jason; Lakonishok, Josef
- NBER Working Paper Series
High-dimensional graphs and variable selection with the Lasso
journal, June 2006
- Meinshausen, Nicolai; Bühlmann, Peter
- The Annals of Statistics, Vol. 34, Issue 3
Lethality and centrality in protein networks
journal, May 2001
- Jeong, H.; Mason, S. P.; Barabási, A. -L.
- Nature, Vol. 411, Issue 6833
Condition-number-regularized covariance estimation
journal, December 2012
- Won, Joong-Ho; Lim, Johan; Kim, Seung-Jean
- Journal of the Royal Statistical Society: Series B (Statistical Methodology), Vol. 75, Issue 3
A Cancer-associated Aurora A Mutant Is Mislocalized and Misregulated Due to Loss of Interaction with TPX2
journal, October 2009
- Bibby, Rachel Ann; Tang, Chan; Faisal, Amir
- Journal of Biological Chemistry, Vol. 284, Issue 48
Statistical Analysis of Non-Lattice Data
journal, September 1975
- Besag, Julian
- The Statistician, Vol. 24, Issue 3
An introduction to ROC analysis
journal, June 2006
- Fawcett, Tom
- Pattern Recognition Letters, Vol. 27, Issue 8
Partial Correlation Estimation by Joint Sparse Regression Models
journal, June 2009
- Peng, Jie; Wang, Pei; Zhou, Nengfeng
- Journal of the American Statistical Association, Vol. 104, Issue 486
Works referencing / citing this record:
Fused lasso regression for identifying differential correlations in brain connectome graphs
journal, July 2018
- Yu, Donghyeon; Lee, Sang Han; Lim, Johan
- Statistical Analysis and Data Mining: The ASA Data Science Journal, Vol. 11, Issue 5
Assisted graphical model for gene expression data analysis
journal, March 2019
- Fan, Xinyan; Fang, Kuangnan; Ma, Shuangge
- Statistics in Medicine, Vol. 38, Issue 13
Robust and sparse Gaussian graphical modelling under cell-wise contamination: Robust graphical modeling
journal, January 2018
- Katayama, Shota; Fujisawa, Hironori; Drton, Mathias
- Stat, Vol. 7, Issue 1
Inferring large graphs using $$\ell _1$$ ℓ 1 -penalized likelihood
journal, August 2017
- Champion, Magali; Picheny, Victor; Vignes, Matthieu
- Statistics and Computing, Vol. 28, Issue 4
Bayesian Discriminant Analysis Using a High Dimensional Predictor
journal, August 2018
- Du, Xingqi; Ghosal, Subhashis
- Sankhya A, Vol. 80, Issue S1
High-dimensional Markowitz portfolio optimization problem: empirical comparison of covariance matrix estimators
journal, February 2019
- Choi, Young-Geun; Lim, Johan; Choi, Sujung
- Journal of Statistical Computation and Simulation, Vol. 89, Issue 7
Graph-Guided Banding of the Covariance Matrix
journal, July 2018
- Bien, Jacob
- Journal of the American Statistical Association, Vol. 114, Issue 526
Foundational principles for large scale inference: Illustrations through correlation mining
preprint, January 2015
- Hero, Alfred O.; Rajaratnam, Bala
- arXiv
The Multiple Quantile Graphical Model
preprint, January 2016
- Ali, Alnur; Kolter, J. Zico; Tibshirani, Ryan J.
- arXiv
A convex framework for high-dimensional sparse Cholesky based covariance estimation
preprint, January 2016
- Khare, Kshitij; Oh, Sang; Rahman, Syed
- arXiv
Endogenous Representation of Asset Returns
preprint, January 2020
- Zhou, Zhipu; Shkolnik, Alexander; Oh, Sang-Yun
- arXiv
Figures / Tables found in this record: