Arief Gusnanto
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Arief Gusnanto has a total of 221 co-authors. Here are the top co-authors by number of shared publications:
- Iain C. Macaulay
Earlham Institute
6 shared publications Frank Dudbridge
Medical Research Council Biostatistics Unit, Cambridge, United Kingdom
5 shared publicationsYudi Pawitan
Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm SE-17177, Sweden
5 shared publications- Alexander Ploner
Karolinska Institutet
4 shared publications - Henry M. Wood
University of Leeds
4 shared publications Philippa Burns
Department of Hematology, University of Cambridge and National Health Service Blood and Transplant, Cambridge, United Kingdom
4 shared publications
Latest Publications
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Biology and Bioinformatics
- Vol. 23 (5) , 2334-2345
Sparse Logistic Regression on Genomic Data for Prediction of Tumour Pathological Subtype
The correct prediction of tumour subtype is critical for the treatment of cancer patients to maximise the chance of survival. The patients' genomic information, such as copy number alterations (CNA) profile, has increasingly become an important factor in the prediction to supplement the traditional pathological subtyping. The incorporation of the CNA information in a prediction model, such as logistic regression, faces two major statistical challenges: first, how to estimate the model parameters in the thousands and, second, how to deal with the correlation of CNA between genomic regions. To address them, we propose a sparse logistic regression model with random effects where some of its parameters are estimated to zero while the other parameters are non-zero. In effect, a variable selection is embedded in the modelling. To deal with the correlation of CNA across genomic regions, we extend further the model to incorporate an additional penalty in the corresponding likelihood function in the logistic regression. The results show that we can identify selected genomic regions that are informative to distinguish different tumour subtypes, while giving a good prediction ability. We illustrate the methodology using CNA dataset from a lung cancer cohort.
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Biology and Bioinformatics
- Vol. 23 (5) , 2334-2345
Sparse Logistic Regression on Genomic Data for Prediction of Tumour Pathological Subtype
The correct prediction of tumour subtype is critical for the treatment of cancer patients to maximise the chance of survival. The patients' genomic information, such as copy number alterations (CNA) profile, has increasingly become an important factor in the prediction to supplement the traditional pathological subtyping. The incorporation of the CNA information in a prediction model, such as logistic regression, faces two major statistical challenges: first, how to estimate the model parameters in the thousands and, second, how to deal with the correlation of CNA between genomic regions. To address them, we propose a sparse logistic regression model with random effects where some of its parameters are estimated to zero while the other parameters are non-zero. In effect, a variable selection is embedded in the modelling. To deal with the correlation of CNA across genomic regions, we extend further the model to incorporate an additional penalty in the corresponding likelihood function in the logistic regression. The results show that we can identify selected genomic regions that are informative to distinguish different tumour subtypes, while giving a good prediction ability. We illustrate the methodology using CNA dataset from a lung cancer cohort.
- 10 June 2026
- Published by American Speech Language Hearing Association in Journal of Speech, Language, and Hearing Research
- Vol. 69 (6) , 2757-2772
Using the Quantifying Bilingual Language Experience Questionnaire to Identify Risk for Language Impairment in Bilingual Children
Purpose: Clinical triage, but also inclusive population sampling in research, requires efficient detection of children likely to be at risk for language impairment (LI), a significant challenge in the case of bilingual children. Our goal was to derive automatic risk factor indices from a parental questionnaire and compare their usefulness in initial identification of children at risk for LI. Method: Alternative risk factor indices were derived from the Quantifying Bilingual Language Experience (Q-BEx) online questionnaire backend calculator, differing in the weight and detail given to language proficiency. Each index was applied to two data sets composed of 5- to 8-year-old children: one with independent diagnosis of developmental language disorder (DLD)/typical development (TD; 109 bilingual children tested in France) and one of largely all-comer children (278 bilingual and monolingual children tested in France, the Netherlands, and the United Kingdom). We compared how these alternative indices predicted (likely) DLD/TD status and structural language outcomes, assessed with Language Impairment Testing in Multilingual Settings tools, designed for bilingual children. Results: RF4-min, a risk factor index composed of four equally weighted components (age of first word, age of first sentence, early development concerns, and strongest speaking skills between the home/heritage language and the societal language), satisfactorily separated bilingual children according to developmental status and was a significant predictor of language outcomes. This composite score showed the best balance between separability, sensitivity, and predictive value compared to indices without a proficiency component or with a more heavily weighted/detailed proficiency component. Conclusion: RF4-min, a Q-BEx automatically derived composite risk factor index, yielded promising results, indicating usefulness as part of first-level triage for LI. Supplemental Material: https://doi.org/10.23641/asha.32080185
- 10 June 2026
- Published by American Speech Language Hearing Association in Journal of Speech, Language, and Hearing Research
- Vol. 69 (6) , 2757-2772
Using the Quantifying Bilingual Language Experience Questionnaire to Identify Risk for Language Impairment in Bilingual Children
Purpose: Clinical triage, but also inclusive population sampling in research, requires efficient detection of children likely to be at risk for language impairment (LI), a significant challenge in the case of bilingual children. Our goal was to derive automatic risk factor indices from a parental questionnaire and compare their usefulness in initial identification of children at risk for LI. Method: Alternative risk factor indices were derived from the Quantifying Bilingual Language Experience (Q-BEx) online questionnaire backend calculator, differing in the weight and detail given to language proficiency. Each index was applied to two data sets composed of 5- to 8-year-old children: one with independent diagnosis of developmental language disorder (DLD)/typical development (TD; 109 bilingual children tested in France) and one of largely all-comer children (278 bilingual and monolingual children tested in France, the Netherlands, and the United Kingdom). We compared how these alternative indices predicted (likely) DLD/TD status and structural language outcomes, assessed with Language Impairment Testing in Multilingual Settings tools, designed for bilingual children. Results: RF4-min, a risk factor index composed of four equally weighted components (age of first word, age of first sentence, early development concerns, and strongest speaking skills between the home/heritage language and the societal language), satisfactorily separated bilingual children according to developmental status and was a significant predictor of language outcomes. This composite score showed the best balance between separability, sensitivity, and predictive value compared to indices without a proficiency component or with a more heavily weighted/detailed proficiency component. Conclusion: RF4-min, a Q-BEx automatically derived composite risk factor index, yielded promising results, indicating usefulness as part of first-level triage for LI. Supplemental Material: https://doi.org/10.23641/asha.32080185
- Open Access
- 5 June 2026
- Published by Elsevier in SSRN Electronic Journal
An Adaptive Fractional Mapping Method for Continuous-Time Optimal Control and Decision Analytics
This paper develops an adaptive fractional-mapped Legendre--Gauss--Radau (LGR) pseudospectral method for continuous-time optimal control problems. The main idea is to introduce a fractional-parameter time mapping that modifies the physical distribution of LGR collocation nodes while preserving the standard collocation structure on the reference interval. The mapping is in corporated into a multi-interval LGR transcription and combined with a ph -mesh refinement strategy, in which the local polynomial degree is increased on smooth intervals and the mesh is subdivided when additional local resolution is required. The continuous Bolza optimal control problem is transcribed into a nonlinear programming problem, and both differential and integration-matrix forms of the collocation constraints are presented. A convergence analysis is established by proving the consistency of the fractional-mapped collocation dynamics, the convergence of the LGR quadrature approximation for the Bolza functional, and the global L2-convergence of the state approximation on the physical time domain under polynomial-degree refinement and mesh refinement. The method is evaluated on the hypersensitive optimal control problem, which has an analytic solution, and the Goddard rocket problem, which represents a nonlinear free-final-time trajectory optimization problem. Numerical results show that the proposed method satisfies the prescribed error tolerance and captures boundary-layer behavior, switching structures, and rapidly varying solution profiles. The case α=1 recovers the classical LGR method, whereas α<1 provides an additional mechanism for node redistribution. The results indicate that the fractional mapping improves the flexibility of adaptive LGR discretization while maintaining accurate objective values.
- 5 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAn Adaptive Fractional Mapping Method for Continuous-Time Optimal Control and Decision Analytics
This paper develops an adaptive fractional-mapped Legendre--Gauss--Radau (LGR) pseudospectral method for continuous-time optimal control problems. The main idea is to introduce a fractional-parameter time mapping that modifies the physical distribution of LGR collocation nodes while preserving the standard collocation structure on the reference interval. The mapping is in corporated into a multi-interval LGR transcription and combined with a ph -mesh refinement strategy, in which the local polynomial degree is increased on smooth intervals and the mesh is subdivided when additional local resolution is required. The continuous Bolza optimal control problem is transcribed into a nonlinear programming problem, and both differential and integration-matrix forms of the collocation constraints are presented. A convergence analysis is established by proving the consistency of the fractional-mapped collocation dynamics, the convergence of the LGR quadrature approximation for the Bolza functional, and the global L2-convergence of the state approximation on the physical time domain under polynomial-degree refinement and mesh refinement. The method is evaluated on the hypersensitive optimal control problem, which has an analytic solution, and the Goddard rocket problem, which represents a nonlinear free-final-time trajectory optimization problem. Numerical results show that the proposed method satisfies the prescribed error tolerance and captures boundary-layer behavior, switching structures, and rapidly varying solution profiles. The case α=1 recovers the classical LGR method, whereas α<1 provides an additional mechanism for node redistribution. The results indicate that the fractional mapping improves the flexibility of adaptive LGR discretization while maintaining accurate objective values.
- 1 April 2026
- Published by Pleiades Publishing Ltd in Lobachevskii Journal of Mathematics
- Vol. 47 (4) , 1683-1696
Factorial-Decay Lag Adaptive LASSO in High-Dimensional ADL Models
This paper proposes the factorial-decay lag adaptive LASSO (FLAdaLASSO), a penalization approach for variable selection in high-dimensional autoregressive time-series models. The method extends existing lag-weighted LASSO procedures by introducing a factorial-decay weighting structure that imposes increasingly strong shrinkage on higher-order lags, thereby enhancing discrimination between relevant and irrelevant predictors in persistent and highly collinear settings. Monte Carlo simulation experiments across a range of sample sizes, lag dimensions, and dependence structures show that FLAdaLASSO improves lag-selection accuracy, reduces overselection, and stabilizes parameter estimation relative to standard LASSO-type methods, while delivering forecasting performance comparable to or better than existing lag-weighted adaptive approaches. Empirical applications to inflation data from Thailand, China, and the United States further illustrate the practical relevance of the proposed method, highlighting its ability to achieve low forecasting error with relatively parsimonious model specifications across heterogeneous macroeconomic environments.
- 1 April 2026
- Published by Pleiades Publishing Ltd in Lobachevskii Journal of Mathematics
- Vol. 47 (4) , 1683-1696
Factorial-Decay Lag Adaptive LASSO in High-Dimensional ADL Models
This paper proposes the factorial-decay lag adaptive LASSO (FLAdaLASSO), a penalization approach for variable selection in high-dimensional autoregressive time-series models. The method extends existing lag-weighted LASSO procedures by introducing a factorial-decay weighting structure that imposes increasingly strong shrinkage on higher-order lags, thereby enhancing discrimination between relevant and irrelevant predictors in persistent and highly collinear settings. Monte Carlo simulation experiments across a range of sample sizes, lag dimensions, and dependence structures show that FLAdaLASSO improves lag-selection accuracy, reduces overselection, and stabilizes parameter estimation relative to standard LASSO-type methods, while delivering forecasting performance comparable to or better than existing lag-weighted adaptive approaches. Empirical applications to inflation data from Thailand, China, and the United States further illustrate the practical relevance of the proposed method, highlighting its ability to achieve low forecasting error with relatively parsimonious model specifications across heterogeneous macroeconomic environments.
- Open Access
- 7 January 2026
- Published by Oxford University Press (OUP) in Bioinformatics Advances
- Vol. 6 (1)
False discovery rate control for sparse correlation matrix
Motivation: In genomic data analysis, high-dimensional correlation matrices pose challenges in interpreting complex relationships among genomic features, particularly due to overfitting. Traditional approaches generate sparse correlation matrices by setting correlations below a predetermined threshold to zero, with the threshold typically determined through Frobenius norm minimization via cross-validation. Results: In this study, we propose a novel alternative thresholding strategy based on an empirical Bayes framework that controls the false discovery rate among nonzero correlations. This method offers two key advantages: first, it is compatible with any correlation measure and, second, it provides an estimation of the expected false positives among nonzero correlations. By formulating the construction of a sparse correlation matrix as both an estimation and an inference problem, our approach enhances interpretability and demonstrates greater sensitivity in retaining true biological correlations than threshold selection based on the minimization of Frobenius norm. Availability and implementation: The relevant R package called scfdr is available from github.com/ariefgusnanto/scfdr
- 7 January 2026
- Published by Oxford University Press (OUP) in Bioinformatics Advances
- Vol. 6 (1)
Open AccessFalse discovery rate control for sparse correlation matrix
Motivation: In genomic data analysis, high-dimensional correlation matrices pose challenges in interpreting complex relationships among genomic features, particularly due to overfitting. Traditional approaches generate sparse correlation matrices by setting correlations below a predetermined threshold to zero, with the threshold typically determined through Frobenius norm minimization via cross-validation. Results: In this study, we propose a novel alternative thresholding strategy based on an empirical Bayes framework that controls the false discovery rate among nonzero correlations. This method offers two key advantages: first, it is compatible with any correlation measure and, second, it provides an estimation of the expected false positives among nonzero correlations. By formulating the construction of a sparse correlation matrix as both an estimation and an inference problem, our approach enhances interpretability and demonstrates greater sensitivity in retaining true biological correlations than threshold selection based on the minimization of Frobenius norm. Availability and implementation: The relevant R package called scfdr is available from github.com/ariefgusnanto/scfdr
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 163446-163477
A Legendre Collocation ph -Mesh Refinement Strategy With Fractional Mapping for Continuous-Time Optimal Control
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 163446-163477
Open AccessA Legendre Collocation ph -Mesh Refinement Strategy With Fractional Mapping for Continuous-Time Optimal Control
- Open Access
- 7 February 2024
- Published by Springer Nature in Genome Biology
- Vol. 25 (1) , 1-29
IDHwt glioblastomas can be stratified by their transcriptional response to standard treatment, with implications for targeted therapy
Background: Glioblastoma (GBM) brain tumors lackingIDH1mutations (IDHwt) have the worst prognosis of all brain neoplasms. Patients receive surgery and chemoradiotherapy but tumors almost always fatally recur.Results: Using RNA sequencing data from 107 pairs of pre- and post-standard treatment locally recurrent IDHwt GBM tumors, we identify two responder subtypes based on longitudinal changes in gene expression. In two thirds of patients, a specific subset of genes is upregulated from primary to recurrence (Up responders), and in one third, the same genes are downregulated (Down responders), specifically in neoplastic cells. Characterization of the responder subtypes indicates subtype-specific adaptive treatment resistance mechanisms that are associated with distinct changes in the tumor microenvironment. In Up responders, recurrent tumors are enriched in quiescent proneural GBM stem cells and differentiated neoplastic cells, with increased interaction with the surrounding normal brain and neurotransmitter signaling, whereas Down responders commonly undergo mesenchymal transition. ChIP-sequencing data from longitudinal GBM tumors suggests that the observed transcriptional reprogramming could be driven by Polycomb-based chromatin remodeling rather than DNA methylation.Conclusions: We show that the responder subtype is cancer-cell intrinsic, recapitulated in in vitro GBM cell models, and influenced by the presence of the tumor microenvironment. Stratifying GBM tumors by responder subtype may lead to more effective treatment.
- 7 February 2024
- Published by Springer Nature in Genome Biology
- Vol. 25 (1) , 1-29
Open AccessIDHwt glioblastomas can be stratified by their transcriptional response to standard treatment, with implications for targeted therapy
Background: Glioblastoma (GBM) brain tumors lackingIDH1mutations (IDHwt) have the worst prognosis of all brain neoplasms. Patients receive surgery and chemoradiotherapy but tumors almost always fatally recur.Results: Using RNA sequencing data from 107 pairs of pre- and post-standard treatment locally recurrent IDHwt GBM tumors, we identify two responder subtypes based on longitudinal changes in gene expression. In two thirds of patients, a specific subset of genes is upregulated from primary to recurrence (Up responders), and in one third, the same genes are downregulated (Down responders), specifically in neoplastic cells. Characterization of the responder subtypes indicates subtype-specific adaptive treatment resistance mechanisms that are associated with distinct changes in the tumor microenvironment. In Up responders, recurrent tumors are enriched in quiescent proneural GBM stem cells and differentiated neoplastic cells, with increased interaction with the surrounding normal brain and neurotransmitter signaling, whereas Down responders commonly undergo mesenchymal transition. ChIP-sequencing data from longitudinal GBM tumors suggests that the observed transcriptional reprogramming could be driven by Polycomb-based chromatin remodeling rather than DNA methylation.Conclusions: We show that the responder subtype is cancer-cell intrinsic, recapitulated in in vitro GBM cell models, and influenced by the presence of the tumor microenvironment. Stratifying GBM tumors by responder subtype may lead to more effective treatment.
- Open Access
- 3 July 2023
- Published by Association of Basic Medical Sciences of FBIH in Biomolecules and Biomedicine
- Vol. 23 (4) , 546-567
Myocarditis and coronavirus disease 2019 vaccination: A systematic review and meta-summary of cases
Vaccination is significant to control, mitigate, and recover from the destructive effects of coronavirus disease 2019 (COVID-19). The incidence of myocarditis following COVID-19 vaccination has been increasing and growing public concern; however, little is known about it. This study aimed to systematically review myocarditis following COVID-19 vaccination. We included studies containing individual patient data of myocarditis following COVID-19 vaccination published between January 1, 2020 and September 7, 2022 and excluded review articles. Joanna Briggs Institute critical appraisals were used for risk of bias assessment. Descriptive and analytic statistics were performed. A total of 121 reports and 43 case series from five databases were included. We identified 396 published cases of myocarditis and observed that the majority of cases was male patients, happened following the second dose of mRNA vaccine administration, and experienced chest pain as a symptom. Previous COVID-19 infection was significantly associated (p < 0.01; OR, 5.74; 95% CI, 2.42–13.64) with the risk of myocarditis following the administration of the first dose, indicating that its primary mechanism is immune-mediated. Moreover, 63 histopathology examinations were dominated by non-infective subtypes. Electrocardiography and cardiac marker combination is a sensitive screening modality. However, cardiac magnetic resonance is a significant noninvasive examination to confirm myocarditis. Endomyocardial biopsy may be considered in confusing and severe cases. Myocarditis following COVID-19 vaccination is relatively benign, with a median length of hospitalization of 5 days, intensive care unit admission of <12%, and mortality of <2%. The majority was treated with nonsteroidal anti-inflammatory drugs, colchicine, and steroids. Surprisingly, deceased cases had characteristics of being female, older age, non-chest pain symptoms, first-dose vaccination, left ventricular ejection fraction of <30%, fulminant myocarditis, and eosinophil infiltrate histopathology.
- 3 July 2023
- Published by Association of Basic Medical Sciences of FBIH in Biomolecules and Biomedicine
- Vol. 23 (4) , 546-567
Open AccessMyocarditis and coronavirus disease 2019 vaccination: A systematic review and meta-summary of cases
Vaccination is significant to control, mitigate, and recover from the destructive effects of coronavirus disease 2019 (COVID-19). The incidence of myocarditis following COVID-19 vaccination has been increasing and growing public concern; however, little is known about it. This study aimed to systematically review myocarditis following COVID-19 vaccination. We included studies containing individual patient data of myocarditis following COVID-19 vaccination published between January 1, 2020 and September 7, 2022 and excluded review articles. Joanna Briggs Institute critical appraisals were used for risk of bias assessment. Descriptive and analytic statistics were performed. A total of 121 reports and 43 case series from five databases were included. We identified 396 published cases of myocarditis and observed that the majority of cases was male patients, happened following the second dose of mRNA vaccine administration, and experienced chest pain as a symptom. Previous COVID-19 infection was significantly associated (p < 0.01; OR, 5.74; 95% CI, 2.42–13.64) with the risk of myocarditis following the administration of the first dose, indicating that its primary mechanism is immune-mediated. Moreover, 63 histopathology examinations were dominated by non-infective subtypes. Electrocardiography and cardiac marker combination is a sensitive screening modality. However, cardiac magnetic resonance is a significant noninvasive examination to confirm myocarditis. Endomyocardial biopsy may be considered in confusing and severe cases. Myocarditis following COVID-19 vaccination is relatively benign, with a median length of hospitalization of 5 days, intensive care unit admission of <12%, and mortality of <2%. The majority was treated with nonsteroidal anti-inflammatory drugs, colchicine, and steroids. Surprisingly, deceased cases had characteristics of being female, older age, non-chest pain symptoms, first-dose vaccination, left ventricular ejection fraction of <30%, fulminant myocarditis, and eosinophil infiltrate histopathology.
- 1 January 2023
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 2913 (1)
Comparison of wavelet methods for segmentation of high-throughput data
High-throughput sequencing (HTS) technology allows sequencing and detection of copy number of DNA segments. Copy number data, in particular, those from HTS technologies are well-known to contain random error variability that makes estimation challenging. Wavelet methods are some of the most powerful methodologies to denoise this data as wavelets, especially Haar wavelet, are well suited for handling the sudden changes seen in HTS data. Many wavelet methods have been proposed to perform segmentation but, in terms of copy number studies, there is no broad comparison of various characteristics of these methods. In this article, we compare three Haar wavelet-based methods for the analysis of HTS data. To make an objective performance assessment, we apply each of these methods to various possible copy number profiles with known truth. The comparison is evaluated in terms of mean integrated square error and Receiver Operating Characteristic (ROC) to measure the accuracy and the operating characteristic of the estimators, respectively. The simulation study indicates that unbalanced Haar wavelet can achieve greater power in estimating the aberrant spot without producing too many spurious change-points compared to the methods use balanced Haar wavelet.
- 1 January 2023
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 2913 (1)
Comparison of wavelet methods for segmentation of high-throughput data
High-throughput sequencing (HTS) technology allows sequencing and detection of copy number of DNA segments. Copy number data, in particular, those from HTS technologies are well-known to contain random error variability that makes estimation challenging. Wavelet methods are some of the most powerful methodologies to denoise this data as wavelets, especially Haar wavelet, are well suited for handling the sudden changes seen in HTS data. Many wavelet methods have been proposed to perform segmentation but, in terms of copy number studies, there is no broad comparison of various characteristics of these methods. In this article, we compare three Haar wavelet-based methods for the analysis of HTS data. To make an objective performance assessment, we apply each of these methods to various possible copy number profiles with known truth. The comparison is evaluated in terms of mean integrated square error and Receiver Operating Characteristic (ROC) to measure the accuracy and the operating characteristic of the estimators, respectively. The simulation study indicates that unbalanced Haar wavelet can achieve greater power in estimating the aberrant spot without producing too many spurious change-points compared to the methods use balanced Haar wavelet.
- Open Access
- 21 September 2022
- Published by Taylor & Francis in Journal of Applied Statistics
- Vol. 51 (2) , 279-297
Model averaging in calibration of near-infrared instruments with correlated high-dimensional data
Model averaging (MA) is a modelling strategy where the uncertainty in the configuration of selected variables is taken into account by weight-combining each estimate of the so-called ‘candidate model’. Some studies have shown that MA enables better prediction, even in high-dimensional cases. However, little is known about the model prediction performance at different types of multicollinearity in high-dimensional data. Motivated by calibration of near-infrared (NIR) instruments,we focus on MA prediction performance in such data. The weighting schemes that we consider are based on the Akaike’s information criterion (AIC), Mallows’ Cp, and cross-validation. For estimating the model parameters, we consider the standard least squares and the ridge regression methods. The results indicate that MA outperforms model selection methods such as LASSO and SCAD in high-correlation data. The use of Mallows’ Cp and cross-validation for the weights tends to yield similar results in all structures of correlation, although the former is generally preferred. We also find that the ridge model averaging outperforms the least-squares model averaging. This research suggests ridge model averaging to build a relatively better prediction of the NIR calibration model.
- 21 September 2022
- Published by Taylor & Francis in Journal of Applied Statistics
- Vol. 51 (2) , 279-297
Open AccessModel averaging in calibration of near-infrared instruments with correlated high-dimensional data
Model averaging (MA) is a modelling strategy where the uncertainty in the configuration of selected variables is taken into account by weight-combining each estimate of the so-called ‘candidate model’. Some studies have shown that MA enables better prediction, even in high-dimensional cases. However, little is known about the model prediction performance at different types of multicollinearity in high-dimensional data. Motivated by calibration of near-infrared (NIR) instruments,we focus on MA prediction performance in such data. The weighting schemes that we consider are based on the Akaike’s information criterion (AIC), Mallows’ Cp, and cross-validation. For estimating the model parameters, we consider the standard least squares and the ridge regression methods. The results indicate that MA outperforms model selection methods such as LASSO and SCAD in high-correlation data. The use of Mallows’ Cp and cross-validation for the weights tends to yield similar results in all structures of correlation, although the former is generally preferred. We also find that the ridge model averaging outperforms the least-squares model averaging. This research suggests ridge model averaging to build a relatively better prediction of the NIR calibration model.