Open research questions in Advanced Statistical Methods and Models

What the literature leaves open

• The need for careful tuning of the regularisation hyperparameter in logistic regression, particularly in high-dimensional data. The lack of a computationally efficient method for probabilistic classification that does not require user-tuned hyperparameters. The need for a model that combines the efficiency of ridge regression with the probabilistic predictions of logistic regression.

  Prevalidated Ridge Regression is a Highly-Efficient Drop-In Replacement for Logistic Regression for High-dimensional Data · 2026 · DOI

• Investigating the finite-sample properties of the subvector Lagrange multiplier test, - Extending the results to non-linear instrumental variables regression models, - Developing new tests that can handle a large number of instruments without an increase in critical values

  Weak-instrument-robust subvector inference in instrumental variables regression: A subvector Lagrange multiplier test and properties of subvector Anderson-Rubin confidence sets · 2026 · DOI

• The lack of weak-instrument-robust subvector tests that recover the degrees of freedom of the standard Wald test. The need for a closed-form solution for subvector confidence sets obtained by inverting the subvector Anderson-Rubin test. The challenge of handling a large number of instruments without an increase in critical values.

  Weak-instrument-robust subvector inference in instrumental variables regression: A subvector Lagrange multiplier test and properties of subvector Anderson-Rubin confidence sets · 2026 · DOI

• Further research is needed to explore the properties of the MCSS estimator. Future studies could investigate the performance of the MCSS estimator in different scenarios.

  The modified conditional sum-of-squares estimator for fractionally integrated models · 2026 · DOI

• The CSS estimator has a biased score, particularly when the data are stationary. The bias of the CSS estimator affects the accuracy of estimates and tests. There is a need for a modified estimator that can address this issue.

  The modified conditional sum-of-squares estimator for fractionally integrated models · 2026 · DOI

• Abstract Although there exists an extensive amount of research on subscores and their properties, limited research has been conducted on categorical subscores and their interpretations.

  A Note on the Use of Categorical Subscores · 2025 · DOI

• Existing methods for addressing multicollinearity in Gaussian linear regression models have limitations and may not always provide reliable results. There is a need for novel hybrid estimators that can improve estimator stability and predictive accuracy.

  Performance of Proposed Ridge – PCA Estimators: Simulation Evidence and Real Data Applications · 2026 · DOI

• Multicollinearity among explanatory variables can lead to unstable parameter estimates in Poisson regression. The choice of the ridge parameter in Poisson ridge regression remains methodologically constrained. Outliers can substantially undermine the reliability and statistical power of empirical analyses.

  Beyond MSE in Poisson Ridge Regression: New Ridge Parameter Estimators with Additional Distributional Performance Criteria · 2026 · DOI

• The existing literature predominantly evaluates ridge parameter estimators using only the mean squared error criterion, neglecting their distributional properties and estimation stability. The choice of the ridge parameter in Poisson ridge regression remains methodologically constrained. There is a need for a multidimensional framework that evaluates ridge parameter estimators beyond mean squared error.

  Beyond MSE in Poisson Ridge Regression: New Ridge Parameter Estimators with Additional Distributional Performance Criteria · 2026 · DOI

• Existing ridge estimators are typically tailored to specific scenarios, limiting their universal applicability. Prior work has not adequately addressed the issue of error variances in the estimation process. There is a need for improved data-driven shrinkage estimators that can effectively address severe multicollinearity.

  Improved Data-Driven Shrinkage Estimators for Regression Models Under Severe Multicollinearity · 2026 · DOI

• Existing methods mainly focus on univariate response settings, and there is a need for methods that can handle multivariate response regression. The problem of heterogeneity in treatment effects is not well addressed by existing methods.

  Simultaneous heterogeneity and reduced-rank learning for multivariate response regression · 2026 · DOI

• Multicollinearity can lead to inaccurate estimation of regression coefficients. The high volume aspect of big data can make it difficult to analyze and interpret the data. The study of multicollinearity in big data analysis is limited by the lack of established methods and techniques.

  Navigating Multicollinearity in Linear Regression Models: Implications for Big Data Analysis · 2026 · DOI

• Future research can explore the application of fixed-order PCA to other domains. Future research can develop methods for consistently estimating r. Future research can explore the use of other rotations for recovering the low-rank signal and factor space.

  Fixed-order PCA: Theory for Overestimated Factor Models · 2026

• Prior work typically requires consistent estimation of the true number of factors r. The paper identifies a gap in the literature by developing asymptotic theory for PCA when R > r.

  Fixed-order PCA: Theory for Overestimated Factor Models · 2026

• The paper is based on empirical evidence and simulations, and a theoretical proof is not offered. The simulations used a limited number of examples and a specific bias value.

  Explicit Form of the Asymptotic Covariance Matrix of the Normalized Within-Stratum Imbalances Following Minimization with Independent Factors with Application to the Log-Rank Test · 2026 · DOI

• The lack of an explicit form of the asymptotic covariance matrix for the normalized within-stratum imbalances. The need for empirical evidence and simulations to support the derivation of the covariance matrix.

  Explicit Form of the Asymptotic Covariance Matrix of the Normalized Within-Stratum Imbalances Following Minimization with Independent Factors with Application to the Log-Rank Test · 2026 · DOI

• Traditional statistical tools are not suitable for non-normally distributed data. There is a need for a robust and reliable method for analyzing non-normally distributed data.

  Assessment of Repeatability and Reproducibility and Robust Regression of Non-normally Distributed Data · 2026 · DOI

• There is a lack of research on methodologies that can effectively address both multicollinearity and heteroscedasticity simultaneously. Existing approaches may not be robust to the joint presence of these challenges.

  Gini-weighted class of ridge estimators in linear regression with heteroscedastic errors · 2026 · DOI

• The paper suggests that future research could investigate the use of the proposed estimator in other areas. The paper suggests that future research could investigate the use of other estimators for the VQC function.

  Vertical quantile comparison functions estimation in location scale families · 2026 · DOI

• The paper identifies a gap in the existing literature for a semiparametric estimator for the VQC function. The paper notes that the existing nonparametric estimators may not be robust to departures from location-scale models.

  Vertical quantile comparison functions estimation in location scale families · 2026 · DOI

• Future research can evaluate the performance of the GMR 4 model on other datasets. Future research can compare the performance of the GMR 4 model with other models.

  Regularized reduced rank regression for mixed predictor and response variables · 2026 · DOI

• The gap is the lack of a model that can handle high-dimensional data with a large number of predictors. The gap is the lack of a model that can provide a sparse and interpretable solution.

  Regularized reduced rank regression for mixed predictor and response variables · 2026 · DOI

• The lack of theory concerning the asymptotic behaviors of model parameter estimators for crossed random effects. The complicated mathematical forms that arise from random effects being crossed.

  Precise asymptotics for linear mixed models with crossed random effects · 2026 · DOI

• Traditional statistical techniques often rely on assumptions that are rarely met in practice. Working with mixed-type data is a challenge, especially in the presence of outlying units or underlying correlation/association structures.

  New distances for mixed-type data able to cope with redundant information · 2026 · DOI

• The application does not include exhaustive options for statistics and graphing. The user needs to download R and RStudio to run the application locally. The application assumes that amplification efficiencies are between 90 -110% and approximately equal between the target and reference genes.

  ProntoPCR: Efficient qPCR Data Analysis Software · 2026 · DOI

Most-cited papers in Advanced Statistical Methods and Models

• Comparative fit indexes in structural models. · Psychological Bulletin · 1990 · 19,458 citations
• Another look at the instrumental variable estimation of error-components models · Journal of Econometrics · 1995 · 14,898 citations
• Significance tests and goodness of fit in the analysis of covariance structures. · Psychological Bulletin · 1980 · 13,503 citations
• A Rationale and Test for the Number of Factors in Factor Analysis · Psychometrika · 1965 · 6,645 citations
• Generalized impulse response analysis in linear multivariate models · Economics Letters · 1998 · 4,814 citations
• Computational Tools for Probing Interactions in Multiple Linear Regression, Multilevel Modeling, and Latent Curve Analysis · Journal of Educational and Behavioral Statistics · 2006 · 4,118 citations
• A Practitioner’s Guide to Cluster-Robust Inference · The Journal of Human Resources · 2015 · 4,080 citations
• Impulse response analysis in nonlinear multivariate models · Journal of Econometrics · 1996 · 4,071 citations
• On Methods in the Analysis of Profile Data · Psychometrika · 1959 · 3,945 citations
• Partial Least Squares Structural Equation Modeling: Rigorous Applications, Better Results and Higher Acceptance · Long Range Planning · 2013 · 3,860 citations

Most recent work

• Statistical Power, Reliability, and Model Assumptions in Multiple Linear Regression: Assessing and Addressing Common Issues · Measurement and Evaluation in Counseling and Development · 2026
• Simultaneous heterogeneity and reduced-rank learning for multivariate response regression · Journal of Multivariate Analysis · 2026
• Scalable and Robust Regression Models for Continuous Proportional Data · Journal of the American Statistical Association · 2026
• On the definition of median for a random interval · Information Sciences · 2026
• Pre-Modeling Diagnostics for Multi-Source Urban Sustainability Indicators: A Five-Stage Framework for Arid and Semi-Arid Cities · Sustainability · 2026
• The ECD-BIE Estimation Theorem in Spectral Form · Studia Geophysica et Geodaetica · 2026
• Base rate utilization in the estimation of violent and criminal risk. · Journal of Threat Assessment and Management · 2026
• Revisiting Bessel’s Correction and the Bias-Variance Tradeoff in Variance Estimation · The American Statistician · 2026
• A Simple Analytic Approach to Regression with or without Interaction Terms: Integrating the Covariance Structure Model and the Delta Method · Psychometrika · 2026
• Robust Estimation of Polyserial Correlation Coefficients: A Density Power Divergence Approach · Psychometrika · 2026

Related topics in Mathematics