☆ 4.5 Article

Fixed and Random Effects Selection by REML and Pathwise Coordinate Optimization

JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS (2013)

Journal

JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS

Volume 22, Issue 2, Pages 341-355

Publisher

AMER STATISTICAL ASSOC

DOI: 10.1080/10618600.2012.681219

Keywords

BIC; LASSO; Mixed-effects models

Categories

Statistics & Probability

Funding

Ministry of Education, Singapore

Ask authors/readers for more resources

Protocol

Community support

Reagent

Community support

Abstract

We propose a two-stage model selection procedure for the linear mixed-effects models. The procedure consists of two steps: First, penalized restricted log-likelihood is used to select the random effects, and this is done by adopting a Newton-type algorithm. Next, the penalized log-likelihood is used to select the fixed effects via pathwise coordinate optimization to improve the computation efficiency. We prove that our procedure has the oracle properties. Both simulation studies and a real data example are carried out to examine finite sample performance of the proposed fixed and random effects selection procedure. Supplementary materials including R code used in this article and proofs for the theorems are available online.

Authors

I am an author on this paper

Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.5

Not enough ratings

Secondary Ratings

Novelty

-

Significance

-

Scientific rigor

-

Rate this paper

Recommended

Article Multidisciplinary Sciences

Study of Bayesian variable selection method on mixed linear regression models

Yong Li, Hefei Liu, Rubing Li

Summary: Variable selection is crucial in statistics, especially in linear regression models for accurate prediction and interpretation. This study introduces the Bayesian adaptive group Lasso method to address variable selection in mixed linear regression models with hidden states and explanatory variables with a grouping structure. The method effectively determines penalty function and parameters, and calculates the specific form of the fully conditional posterior distribution for each parameter. Simulation experiments and application analysis on Alzheimer's Disease dataset demonstrate its effectiveness in identifying observations from different hidden states, though variable selection results differ across states.

PLOS ONE (2023)

Add to Collection

Article Statistics & Probability

Inference for High-Dimensional Linear Mixed-Effects Models: A Quasi-Likelihood Approach

Sai Li, T. Tony Cai, Hongzhe Li

Summary: The study introduces a quasi-likelihood approach for estimating unknown parameters in linear mixed-effects models, which does not rely on the structural information of the variance components, while also exploring the estimation of variance components with high-dimensional fixed effects.

JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION (2022)

Add to Collection

Article Mathematics

Modified BIC Criterion for Model Selection in Linear Mixed Models

Hang Lai, Xin Gao

Summary: Linear mixed-effects models are widely used in analyzing clustered, hierarchical, and longitudinal data. Model selection in these models is challenging due to the inclusion of both fixed effects and variance component parameters. This article proposes a modified BIC for model selection in linear mixed-effects models that can handle cases where the variance components are on the boundary of the parameter space. Simulation results show that the modified BIC outperforms the regular BIC in most cases, and it is also successfully applied to a real dataset for model selection.

MATHEMATICS (2023)

Add to Collection

Article Physics, Multidisciplinary

Bayesian Variable Selection and Estimation in Semiparametric Simplex Mixed-Effects Models with Longitudinal Proportional Data

Anmin Tang, Xingde Duan, Yuanying Zhao

Summary: In the development of simplex mixed-effects models, random effects are typically assumed to follow a normal distribution. To address violations of this assumption, a centered Dirichlet process mixture model is employed in this paper. By utilizing a Bayesian Lasso, important covariates with nonzero effects can be selected while estimating unknown parameters in semiparametric simplex mixed-effects models, with the help of a block Gibbs sampler and the Metropolis-Hastings algorithm.

ENTROPY (2022)

Add to Collection

Article Biochemical Research Methods

Efficient penalized generalized linear mixed models for variable selection and genetic risk prediction in high-dimensional data

Julien St-Pierre, Karim Oualkacha, Sahir Rai Bhatnagar

Summary: This study introduces a new method called pglmm, a penalized generalized linear mixed model that allows simultaneous selection of genetic markers and estimation of their effects, taking into account between-individual correlations and binary traits. Simulation and data analysis results show that pglmm outperforms other methods in terms of predictive performance in high-dimensional settings.

BIOINFORMATICS (2023)

Add to Collection

Article Social Sciences, Mathematical Methods

Recovering Crossed Random Effects in Mixed-Effects Models Using Model Averaging

Jose Angel Martinez-Huertas, Ricardo Olmos

Summary: The study found that using model averaging, especially Akaike weights, can effectively recover the variances of random effects under certain sample cluster conditions, and therefore recommends using model averaging in MEMs-CR to address model uncertainty.

METHODOLOGY-EUROPEAN JOURNAL OF RESEARCH METHODS FOR THE BEHAVIORAL AND SOCIAL SCIENCES (2022)

Add to Collection

Article Multidisciplinary Sciences

Double Penalized Expectile Regression for Linear Mixed Effects Model

Sihan Gao, Jiaqing Chen, Zihao Yuan, Jie Liu, Yangxin Huang

Summary: This paper introduces a double penalized expectile regression method for linear mixed effects model, which estimates coefficients and selects variables for random and fixed effects simultaneously. The proposed method combines double penalized expectile regression with the linear mixed effects model, and utilizes the iterative Lasso expectile regression algorithm to solve the parameters. The effectiveness of the method is demonstrated through simulation studies and a case study using real data, showing that it is robust to various error distributions and capable of excluding inactive variables and selecting important variables in high-dimensional data.

SYMMETRY-BASEL (2022)

Add to Collection

Article Mathematics

Variational Bayesian Inference in High-Dimensional Linear Mixed Models

Jieyi Yi, Niansheng Tang

Summary: In this paper, a variational Bayesian approach is proposed to simultaneously select variables and estimate parameters in high-dimensional linear mixed models. Compared to traditional methods, this approach is more efficient and flexible. Simulation studies and an empirical analysis demonstrate the performance and practicality of the proposed method.

MATHEMATICS (2022)

Add to Collection

Article Statistics & Probability

Mixed Lasso estimator for stochastic restricted regression models

Huseyin Guler, Ebru Ozgur Guler

Summary: The study proposes a Mixed Lasso (M-Lasso) estimator to simultaneously select the true model and estimate parameters in large datasets, outperforming existing estimators in terms of mean squared error and model selection performance. Applied to estimating parameters of a production function with stochastic restrictions, M-Lasso provides reasonable and more precise estimates aligning with economic theory.

JOURNAL OF APPLIED STATISTICS (2021)

Add to Collection

Article Statistics & Probability

THE BAYESIAN NESTED LASSO FOR MIXED FREQUENCY REGRESSION MODELS

Satyajit Ghosh, Kshitij Khare, George Michailidis

Summary: This paper proposes a Bayesian framework for jointly estimating the appropriate lag and regression coefficients in linear models. By introducing the Bayesian nested lasso (BNL) prior distribution, the framework allows for effective selection and estimation of model parameters, and incorporates desirable decay patterns over time lags in the magnitude of the regression coefficients.

ANNALS OF APPLIED STATISTICS (2023)

Add to Collection

Article Engineering, Environmental

Kinetic parameter estimation with nonlinear mixed-effects models

Thomas Krumpolc, D. W. Trahan, D. A. Hickman, L. T. Biegler

Summary: Applications of fixed-effects models for kinetic parameter estimation assume independence among batches, but biased residuals often exist in multiple longitudinal batch experiments with time series data. Nonlinear mixed-effects models provide an alternative approach to address the two types of random experimental variation resulting from longitudinal experiments: measurement error for each data point and random batch-to-batch variation. In our case study, implementing a mixed-effects model using nonlinear programming for a batch reactor system yields parameter estimates with less bias compared to a fixed-effects model. Additionally, the Bayesian notion of probability shares is applied to discriminate between several candidate mixed-effects models, demonstrating the ability to elucidate additional model information when fixed-effects models are inappropriate.

CHEMICAL ENGINEERING JOURNAL (2022)

Add to Collection

Article Statistics & Probability

LOCALLY ASSOCIATED GRAPHICAL MODELS AND MIXED CONVEX EXPONENTIAL FAMILIES

Steffen Lauritzen, Piotr Zwiernik

Summary: In this paper, a concept of local association is proposed, where highly connected components in a graphical model are positively associated and its properties are studied. The main motivation comes from gene expression data, and the models are instances of mixed convex exponential families. The positive graphical lasso is introduced to relax the positivity assumption by penalizing negative partial correlations, and a GOLAZO algorithm based on block-coordinate descent is developed for optimization problems in graphical models.

ANNALS OF STATISTICS (2022)

Add to Collection

Article Statistics & Probability

Structural inference of time-varying mixed graphical models

Qingyang Liu, Yuping Zhang, Zhengqing Ouyang

Summary: This paper focuses on joint structural estimation of time-varying mixed graphical models based on multivariate data over a series of time points. Various techniques, such as flexible local estimator, group lasso penalty, and an accelerated ADMM-based algorithm, are employed to handle the changing network structure, demonstrating practical merits through synthetic and real data applications.

STAT (2021)

Add to Collection

Review Genetics & Heredity

Penalized linear mixed models for structured genetic data

Anna C. Reisetter, Patrick Breheny

Summary: Many genetic studies aiming to identify genetic variants associated with complex phenotypes are hindered by unobserved confounding factors due to environmental heterogeneity. Penalized linear mixed models (LMMs) are proposed as an effective method to correct for these confounding factors, by modeling relatedness and population structure as random effects with covariance structure estimated from observed genetic data. Despite a wealth of literature on penalized regression and LMMs separately, there is limited discussion on their integration in the context of genetic associations. The study aims to explore the statistical properties of penalized LMMs and their ability to accurately estimate genetic effects in the presence of environmental confounding.

GENETIC EPIDEMIOLOGY (2021)

Add to Collection

Article Environmental Sciences

An assessment of meteorological parameters effects on COVID-19 pandemic in Bangladesh using machine learning models

Jaionto Karmokar, Mohammad Aminul Islam, Machbah Uddin, Md Rakib Hassan, Md Sayeed Iftekhar Yousuf

Summary: This study examined the impact of meteorological parameters on COVID-19 transmission in Bangladesh and used a combination of Random Forest, CART, and Lasso feature selection techniques to analyze their actual effects. The results revealed that minimum temperature and cloud cover are significant factors influencing COVID-19, while wind speed and air quality have a negative impact.

ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH (2022)

Add to Collection

Article Statistics & Probability

Adaptive testing for the partially linear single-index model with error-prone linear covariates

Zhensheng Huang, Quanxi Shao, Zhen Pang, Bingqing Lin

STATISTICAL METHODOLOGY (2015)

Add to Collection

Article Statistics & Probability

REGULARISATION PARAMETER SELECTION VIA BOOTSTRAPPING

Zhen Pang, Bingqing Lin, Jiming Jiang

AUSTRALIAN & NEW ZEALAND JOURNAL OF STATISTICS (2016)

Add to Collection

Article Computer Science, Theory & Methods

Stable prediction in high-dimensional linear models

Bingqing Lin, Qihua Wang, Jun Zhang, Zhen Pang

STATISTICS AND COMPUTING (2017)

Add to Collection

Article Computer Science, Theory & Methods

Large-scale kernel methods for independence testing

Qinyi Zhang, Sarah Filippi, Arthur Gretton, Dino Sejdinovic

STATISTICS AND COMPUTING (2018)

Add to Collection

Article Physics, Mathematical

Cluster feature selection in high-dimensional linear models

Bingqing Lin, Zhen Pang, Qihua Wang

RANDOM MATRICES-THEORY AND APPLICATIONS (2018)

Add to Collection

Article Computer Science, Interdisciplinary Applications

Estimation for the single-index models with random effects

Zhen Pang, Liugen Xue

COMPUTATIONAL STATISTICS & DATA ANALYSIS (2012)

Add to Collection

Article Computer Science, Interdisciplinary Applications

Testing structural change in partially linear single-index models with error-prone linear covariates

Zhensheng Huang, Zhen Pang, Tao Hu

COMPUTATIONAL STATISTICS & DATA ANALYSIS (2013)

Add to Collection

Article Computer Science, Interdisciplinary Applications

Adaptive profile-empirical-likelihood inferences for generalized single-index models

Zhensheng Huang, Zhen Pang, Riquan Zhang

COMPUTATIONAL STATISTICS & DATA ANALYSIS (2013)

Add to Collection

Article Statistics & Probability

Tilted Correlation Screening Learning in High-Dimensional Data Analysis

Bingqing Lin, Zhen Pang

JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS (2014)

Add to Collection

Article Statistics & Probability

Corrected empirical likelihood inference for right-censored partially linear single-index model

Zhensheng Huang, Zhen Pang

JOURNAL OF MULTIVARIATE ANALYSIS (2012)

Add to Collection

Article Statistics & Probability

Efficient penalized estimating method in the partially varying-coefficient single-index model

Zhensheng Huang, Bingqing Lin, Fan Feng, Zhen Pang

JOURNAL OF MULTIVARIATE ANALYSIS (2013)

Add to Collection

Article Statistics & Probability

Model structure selection in single-index-coefficient regression models

Zhensheng Huang, Zhen Pang, Bingqing Lin, Quanxi Shao

JOURNAL OF MULTIVARIATE ANALYSIS (2014)

Add to Collection

Article Computer Science, Theory & Methods

Statistical inference for a single-index varying-coefficient model

Liugen Xue, Zhen Pang

STATISTICS AND COMPUTING (2013)

Add to Collection

Article Biotechnology & Applied Microbiology

Stability of methods for differential expression analysis of RNA-seq data

Bingqing Lin, Zhen Pang

BMC GENOMICS (2019)

Add to Collection

No Data Available

© Peeref 2019-2024. All rights reserved.