Estimating the Optimal Linear Combination of Biomarkers using Spherically Constrained Optimization

Bibhas Chakraborty; Christine B. Peterson; Clifton D. Fuller; Debsurya De; Katherine A. Hutcheson; Mona Kamal; Priyam Das; Raju Maiti

arxiv: 1909.04024 · v2 · submitted 2019-09-07 · 📊 stat.ME · stat.CO

Estimating the Optimal Linear Combination of Biomarkers using Spherically Constrained Optimization

Priyam Das , Debsurya De , Raju Maiti , Mona Kamal , Katherine A. Hutcheson , Clifton D. Fuller , Bibhas Chakraborty , Christine B. Peterson This is my paper

Pith reviewed 2026-05-24 15:20 UTC · model grok-4.3

classification 📊 stat.ME stat.CO

keywords hypervolume under the manifoldpattern searchlinear combination of biomarkersmulti-category classificationglobal optimizationROC generalizationbiomarker weighting

0 comments

The pith

Pattern search optimization finds better linear combinations of biomarkers by maximizing the empirical hypervolume under the manifold for multi-category outcomes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a way to combine continuous predictors optimally when the outcome has more than two categories. This requires maximizing an empirical estimate of the hypervolume under the manifold, the natural extension of the AUC. The objective is discontinuous and multi-modal, so the authors apply a derivative-free pattern search algorithm. Simulations show the method recovers superior coefficient vectors compared with the step-down algorithm, and the technique is used to predict swallowing difficulty after head-and-neck radiation therapy.

Core claim

The pattern search procedure estimates the coefficient vector that maximizes the empirical hypervolume under the manifold more reliably than existing global optimizers, as demonstrated by simulation studies that vary the number of predictors and outcome categories.

What carries the argument

Pattern search algorithm applied to maximization of the discontinuous empirical HUM objective under a spherical constraint on the coefficient vector.

Load-bearing premise

Pattern search will reliably reach the global maximum of the HUM objective without excessive cost or entrapment in local optima as the number of predictors and categories increases.

What would settle it

A simulation study with known true optimal coefficients, increasing numbers of predictors, and multiple outcome categories in which the pattern search method fails to recover the true vector or requires prohibitive runtime.

Figures

Figures reproduced from arXiv: 1909.04024 by Bibhas Chakraborty, Christine B. Peterson, Clifton D. Fuller, Debsurya De, Katherine A. Hutcheson, Mona Kamal, Priyam Das, Raju Maiti.

**Figure 1.** Figure 1: Flowchart of the SCOR algorithm where s denotes the step size, φ denotes the step size threshold, and SOL(k) denotes the solution returned by the k-th run [PITH_FULL_IMAGE:figures/full_fig_p027_1.png] view at source ↗

**Figure 2.** Figure 2: Exploratory moves on the surface of unit sphere by SCOR. Figure (a) shows an initial point (β1, β2, β3) (green) on the surface of the unit sphere and a point explored (β1 + t2, β2 + s, β3 + t2) (red) after making exploratory move with step-size s for the 2nd coordinate, where t2 denotes the adjustment step-size. Figure (b) shows exploratory moves in a typical iteration of SCOR: starting from the current so… view at source ↗

**Figure 3.** Figure 3: Boxplots of the optimal combination vector scores are shown across the outcome categories 0, 1, and 2-4. The methods compared are SCOR ((a) ULBA, (e) EHUM), NM ((b) ULBA, (f) EHUM), Step-down ((c) ULBA, (g) EHUM) and Min-max ((d) ULBA, (h) EHUM). The horizontal dotted lines denote the corresponding cut-points for classification obtained by maximizing Youden’s Index. Since this example includes 3 outcome ca… view at source ↗

read the original abstract

In the context of a binary classification problem, the optimal linear combination of continuous predictors can be estimated by maximizing an empirical estimate of the area under the receiver operating characteristic (ROC) curve (AUC). For multi-category responses, the optimal predictor combination can similarly be obtained by maximization of the empirical hypervolume under the manifold (HUM). This problem is particularly relevant to medical research, where it may be of interest to diagnose a disease with various subtypes or predict a multi-category outcome. Since the empirical HUM is discontinuous, non-differentiable, and possibly multi-modal, solving this maximization problem requires a global optimization technique. Estimation of the optimal coefficient vector using existing global optimization techniques is computationally expensive, becoming prohibitive as the number of predictors and the number of outcome categories increases. We propose an efficient derivative-free black-box optimization technique based on pattern search to solve this problem. Through extensive simulation studies, we demonstrate that the proposed method achieves better performance compared to existing methods including the step-down algorithm. Finally, we illustrate the proposed method to predict swallowing difficulty after radiation therapy for oropharyngeal cancer based on radiation dose to various structures in the head and neck.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Pattern search gives a faster practical option for multi-class HUM maximization but lacks global guarantees and the simulation evidence needs checking for local-optima issues.

read the letter

The paper's main move is to adapt pattern search, a derivative-free direct-search method, to maximize the empirical HUM when combining continuous biomarkers for multi-category outcomes. It frames this as cheaper than standard global optimizers while beating the step-down algorithm on the simulations they ran, and it shows the idea on a head-and-neck cancer dose-toxicity example. That is the concrete contribution: a black-box procedure that sidesteps gradients on a discontinuous, non-smooth objective that matters in diagnostic work. The application example is straightforward and relevant, and the motivation about scaling with more predictors or more categories is clear. The simulations are presented as extensive and supportive of better performance. The soft spot is exactly the one the stress-test note flags. Pattern search is a local method; the paper itself notes the empirical HUM can be multi-modal, yet offers no convergence guarantee to the global maximum and no description here of how many random restarts were used or how the test cases were chosen to expose entrapment. If the simulation design did not include regimes where local optima are common, the performance edge could shrink or disappear as dimension or number of categories grows. The citation pattern looks standard for this niche and does not appear circular. This is the kind of incremental computational tool that medical statisticians who actually fit these models might try out. It is worth sending to referees so they can inspect the simulation protocol and any restart strategy in detail; the work is coherent on its own terms even if the optimality claim rests on empirical evidence that needs scrutiny.

Referee Report

2 major / 1 minor

Summary. The paper proposes a pattern search algorithm for maximizing the empirical hypervolume under the manifold (HUM) to estimate optimal linear combinations of continuous predictors for multi-category outcomes. It positions the method as a computationally efficient derivative-free alternative to existing global optimizers, claims superior performance over the step-down algorithm via extensive simulations, and illustrates the approach on radiation dose data for predicting post-therapy swallowing difficulty.

Significance. If the performance claims hold under rigorous verification of global optimality, the work would supply a practical tool for biomarker combination in multi-class medical diagnostics, where the non-smooth HUM objective becomes prohibitive for standard global methods as the number of predictors and categories grows.

major comments (2)

[Abstract] The abstract states that 'through extensive simulation studies, the proposed method achieves better performance compared to existing methods including the step-down algorithm,' yet supplies no information on the number of replicates, range of p and K, number of random restarts, or quantitative metrics (e.g., achieved HUM values with variability) used to confirm that pattern search located the global rather than a local maximum of the multi-modal empirical HUM.
[Method / Optimization section] The manuscript explicitly notes that the empirical HUM is 'discontinuous, non-differentiable, and possibly multi-modal' and that pattern search is a local direct-search method, but provides neither theoretical convergence guarantees to the global optimum nor empirical evidence (e.g., comparison against a known global solver on small instances) that the algorithm reliably recovers the global maximizer as dimension increases.

minor comments (1)

[Introduction / Methods] Notation for the spherical constraint and the precise definition of the empirical HUM estimator should be stated explicitly in the main text rather than deferred to supplementary material.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

Referee: [Abstract] The abstract states that 'through extensive simulation studies, the proposed method achieves better performance compared to existing methods including the step-down algorithm,' yet supplies no information on the number of replicates, range of p and K, number of random restarts, or quantitative metrics (e.g., achieved HUM values with variability) used to confirm that pattern search located the global rather than a local maximum of the multi-modal empirical HUM.

Authors: We agree that the abstract would benefit from additional detail on the simulation design. The main text (Section 4) already reports the number of replicates, ranges of p and K, use of multiple random restarts, and quantitative HUM metrics with variability. In the revised manuscript we will expand the abstract to briefly summarize these elements so that the performance claims are placed in clearer context. revision: yes
Referee: [Method / Optimization section] The manuscript explicitly notes that the empirical HUM is 'discontinuous, non-differentiable, and possibly multi-modal' and that pattern search is a local direct-search method, but provides neither theoretical convergence guarantees to the global optimum nor empirical evidence (e.g., comparison against a known global solver on small instances) that the algorithm reliably recovers the global maximizer as dimension increases.

Authors: We acknowledge that the manuscript supplies neither theoretical convergence guarantees (which are unavailable for pattern search on this class of objectives) nor direct comparisons against a certified global solver on small instances. The reported simulations compare the method to the step-down algorithm and other heuristics, showing improved average HUM values. In the revision we will add an explicit discussion paragraph noting these limitations and recommending multiple random starts in practice; we do not claim guaranteed global optimality. revision: partial

Circularity Check

0 steps flagged

No circularity: algorithmic proposal validated by independent simulations

full rationale

The paper introduces pattern search as a derivative-free optimizer for the empirical HUM objective and validates its performance via simulation studies against the step-down algorithm and other baselines. No derivation chain exists that reduces a claimed result to its own inputs by construction; the method is presented as an independent algorithmic choice whose superiority is assessed empirically rather than through self-referential fitting or self-citation load-bearing. The optimization procedure and the simulation-based performance claims remain separate, with no equations or premises that define the target quantity in terms of the proposed solution itself.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract; no explicit free parameters, axioms, or invented entities are described.

pith-pipeline@v0.9.0 · 5758 in / 1026 out tokens · 21506 ms · 2026-05-24T15:20:12.795692+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · 2 internal anchors

[1]

and Tolle, J

Boggs, P. and Tolle, J. (1996). Sequential quadratic programming s. Acta Numerica 1–52

work page 1996
[2]

E., Le, Q

Daly, M. E., Le, Q. T., Maxim, P. G., Loo Jr, B., Kaplan, M. J., et al. (2010 ). Intensity- modulated radiotherapy in the treatment of oropharyngeal canc er: clinical outcomes and patterns of failure. Int J Radiat Oncol Biol Phys 76, 1339–1346

work page 2010
[3]

Das, P. (2016a). Black-box optimization on hyper-rectangle using recursive modiﬁed pattern search and application to matrix completion problem with non-convex regularization. arXiv, arxiv.org/pdf/1604.08616.pdf

work page arXiv
[4]

Das, P. (2016b). Black-box optimization on multiple simplex constrain ed blocks. arXiv, arxiv.org/abs/1609.02249

work page internal anchor Pith review Pith/arXiv arXiv
[5]

Das, P. (2020). Recursive modiﬁed pattern search on high-dimens ional simplex : A blackbox optimization technique. The Indian Journal of Statistics: Series B . Duprez et al. (2013). Systematic review of dose–volume correlate s for structures related to late swallowing disturbances after radiotherapy for head and neck cancer. Dysphagia 28, 337–349. Eisbruc...

work page 2020
[6]

and Metropolis, N

Fermi, E. and Metropolis, N. (1952). Numerical solution of a minimum p roblem. los alamos unclassiﬁed report la–1492. Los Alamos National Laboratory, Los Alamos, USA

work page 1952
[7]

Fraser, A. (1957). Simulation of genetic systems by automatic digit al computers i. introduc- tion. Australian Journal of Biological Sciences 10, 484–491

work page 1957
[8]

Geris, L. (2012). Computational Modeling in Tissue Engineering . Springer

work page 2012
[9]

P., Lewin, J

Goepfert, R. P., Lewin, J. S., Barrow, M. P., Warneke, C. L., Fuller, C . D., Lai, S. Y., et al. (2018). Grading dysphagia as a toxicity of head and neck canc er: diﬀerences in severity classiﬁcation based on MBS DIGEST and clinical CTCAE grade s. Dysphagia 33, 185–191

work page 2018
[10]

Horst, R. (2002). Introduction to Global Optimization (Nonconvex Optimizat ion and Its Applications). Springer-Verlag Berlin, Heidelberg

work page 2002
[11]

and Chen, Y

Hsu, M. and Chen, Y. (2016). Optimal linear combination of biomarke rs for multi-category diagnosis. Statistics in Medicine 35, 202–213. Hutcheson et al. (2014). Long-term functional and survival out comes after induction chemotherapy and risk-based deﬁnitive therapy for locally advanc ed squamous cell carcinoma of the head and neck. Head Neck. 36, 474–48...

work page 2016
[12]

S., Volpe, S., Zaveri, J., Barrow, M

Kamal, M., Mohamed, A. S., Volpe, S., Zaveri, J., Barrow, M. P., Gunn, G . B., et al. (2018). Radiotherapy dose–volume parameters predict videoﬂuoroscopy -detected dysphagia per DIGEST after IMRT for oropharyngeal cancer: Results of a pros pective registry. Radio- therapy and Oncology 128, 442–451

work page 2018
[13]

Kang, L., Xiong, C., Crane, P., and Tian, L. (2013). Linear combinatio ns of biomarkers to improve diagnostic accuracy with three ordinal diagnostic categ ories. Statistics in Medicine 32, 631–643. 24 Biometrics,

work page 2013
[14]

Karmakar, N. (1984). New polynomial-time algorithm for linear progr amming. Combina- torica 4, 373–395

work page 1984
[15]

Kerr, C., Dura-Bernal, S., Smolinski, T., Chadderdon, G., and Wilson, D . (2018). Optimiza- tion by adaptive stochastic descent. PLOS One 13, e0192944

work page 2018
[16]

Kirkpatrick, S., Gelatt, C., and Vecchi, M. (1983). Optimization by sim ulated annealing. Australian Journal of Biological Sciences 220, 671–680

work page 1983
[17]

C., Teguh, D

Levendag, P. C., Teguh, D. N., Voet, P., van der Est, H., Noever, I., de Kruijf, W. J., et al. (2007). Dysphagia disorders in patients with cancer of the oropha rynx are signiﬁcantly aﬀected by the radiation therapy dose to the superior and middle co nstrictor muscle: a dose-eﬀect relationship. Radiotherapy and Oncology 85, 64–73

work page 2007
[18]

and Fine, J

Li, J. and Fine, J. (2008). ROC analysis with multiple classes and multiple tests: methodology and its application in microarray studies. Biostatistics 9, 566–576

work page 2008
[19]

Liu, C., Liu, A., and Halabi, S. (2011). A min–max combination of biomark ers to improve diagnostic accuracy. Statistics in Medicine 30, 2005–2014

work page 2011
[20]

and Xiong, C

Luo, J. and Xiong, C. (2012). Diagtest3grp: An r package for ana lyzing diagnostic tests with three ordinal groups. J Stat Softw. 51, 1–24

work page 2012
[21]

and Huang, J

Ma, S. and Huang, J. (2007). Combining multiple markers for classiﬁc ation using ROC. Biometrics 63, 751–757

work page 2007
[22]

Maiti, R., Li, J., Das, P., Feng, L., Hausenloy, D., and Chakraborty, B. (2019). A distribution- free smoothed combination method of biomarkers to improve diagno stic accuracy in multi-category classiﬁcation. arxiv.org/abs/1904.10046

work page internal anchor Pith review Pith/arXiv arXiv 2019
[23]

Mossman, D. (1999). Three-way ROCs. Medical Decision Making 19, 78–89

work page 1999
[24]

and Yiannoutsos, C

Nakas, C. and Yiannoutsos, C. (2004). Ordered multiple-class ROC analysis with continuous measurements. Statistics in Medicine 23, 3437–3449

work page 2004
[25]

and Mead, R

Nelder, J. and Mead, R. (1965). A simplex method for function minimiz ation. Computer Estimating the optimal linear combination of predictors us ing SCOR 25 Journal 7, 308–313

work page 1965
[26]

Pepe, M., Cai, T., and Longton, G. (2006). Combining predictors for classiﬁcation using the area under the receiver operating characteristic curve. Biometrics 62, 221–229

work page 2006
[27]

and Thompson, M

Pepe, M. and Thompson, M. (2000). Combining diagnostic test resu lts to increase accuracy. Biostatistics 1, 123–140. Scurﬁeld, B. (1996). Multiple-event forced-choice tasks in the th eory of signal detectability. Journal of Mathematical Psychology 40, 253–269

work page 2000
[28]

and Liu, J

Su, J. and Liu, J. (1993). Linear combinations of multiple diagnostic m arkers. Journal of the American Statistical Association 88, 1350–1355

work page 1993
[29]

G., Eisenhauer, E

Therasse, P., Arbuck, S. G., Eisenhauer, E. A., Wanders, J., Kaplan , R. S., Rubinstein, L., et al. (2000). New guidelines to evaluate the response to treatmen t in solid tumors. Journal of the National Cancer Institute 92, 205–216

work page 2000
[30]

Torczon, V. (1997). On the convergence of pattern search algo rithms. SIAM Journal on Optimization 7, 1–25

work page 1997
[31]

D., Setser, A., Rusch, V., Jaques, D., Budach , V., et al

Trotti, A., Colevas, A. D., Setser, A., Rusch, V., Jaques, D., Budach , V., et al. (2003). Ctcae v3.0: development of a comprehensive grading system for the adve rse eﬀects of cancer treatment. In Seminars in Radiation Oncology , volume 13, 176–181

work page 2003
[32]

Yanyu, Z. (2010). ROC analysis in diagnostic medicine (phd thesis). Department of Statistics and Applied Probability, National University of Singapore

work page 2010
[33]

Youden, W. (1950). Index for rating diagnostic tests. Cancer 3, 32–35

work page 1950
[34]

and Li, J

Zhang, Y. and Li, J. (2011). Combining multiple markers for multi-cat egory classiﬁcation: An ROC surface approach. Australian & New Zealand Journal of Statistics 53, 63–78

work page 2011
[35]

Zhou, X.-H., McClish, D., and Obuchowski, N. (2002). Statistical Methods in Diagnostic Medicine, volume 464. John Wiley and Sons. Received . . . 2020. Revised . . . 2020. Accepted . . . 2020. 26 Biometrics, . . . . . . Figure 1. Flowchart of the SCOR algorithm where s denotes the step size, φ denotes the step size threshold, and SOL(k) denotes the solutio...

work page 2002

[1] [1]

and Tolle, J

Boggs, P. and Tolle, J. (1996). Sequential quadratic programming s. Acta Numerica 1–52

work page 1996

[2] [2]

E., Le, Q

Daly, M. E., Le, Q. T., Maxim, P. G., Loo Jr, B., Kaplan, M. J., et al. (2010 ). Intensity- modulated radiotherapy in the treatment of oropharyngeal canc er: clinical outcomes and patterns of failure. Int J Radiat Oncol Biol Phys 76, 1339–1346

work page 2010

[3] [3]

Das, P. (2016a). Black-box optimization on hyper-rectangle using recursive modiﬁed pattern search and application to matrix completion problem with non-convex regularization. arXiv, arxiv.org/pdf/1604.08616.pdf

work page arXiv

[4] [4]

Das, P. (2016b). Black-box optimization on multiple simplex constrain ed blocks. arXiv, arxiv.org/abs/1609.02249

work page internal anchor Pith review Pith/arXiv arXiv

[5] [5]

Das, P. (2020). Recursive modiﬁed pattern search on high-dimens ional simplex : A blackbox optimization technique. The Indian Journal of Statistics: Series B . Duprez et al. (2013). Systematic review of dose–volume correlate s for structures related to late swallowing disturbances after radiotherapy for head and neck cancer. Dysphagia 28, 337–349. Eisbruc...

work page 2020

[6] [6]

and Metropolis, N

Fermi, E. and Metropolis, N. (1952). Numerical solution of a minimum p roblem. los alamos unclassiﬁed report la–1492. Los Alamos National Laboratory, Los Alamos, USA

work page 1952

[7] [7]

Fraser, A. (1957). Simulation of genetic systems by automatic digit al computers i. introduc- tion. Australian Journal of Biological Sciences 10, 484–491

work page 1957

[8] [8]

Geris, L. (2012). Computational Modeling in Tissue Engineering . Springer

work page 2012

[9] [9]

P., Lewin, J

Goepfert, R. P., Lewin, J. S., Barrow, M. P., Warneke, C. L., Fuller, C . D., Lai, S. Y., et al. (2018). Grading dysphagia as a toxicity of head and neck canc er: diﬀerences in severity classiﬁcation based on MBS DIGEST and clinical CTCAE grade s. Dysphagia 33, 185–191

work page 2018

[10] [10]

Horst, R. (2002). Introduction to Global Optimization (Nonconvex Optimizat ion and Its Applications). Springer-Verlag Berlin, Heidelberg

work page 2002

[11] [11]

and Chen, Y

Hsu, M. and Chen, Y. (2016). Optimal linear combination of biomarke rs for multi-category diagnosis. Statistics in Medicine 35, 202–213. Hutcheson et al. (2014). Long-term functional and survival out comes after induction chemotherapy and risk-based deﬁnitive therapy for locally advanc ed squamous cell carcinoma of the head and neck. Head Neck. 36, 474–48...

work page 2016

[12] [12]

S., Volpe, S., Zaveri, J., Barrow, M

Kamal, M., Mohamed, A. S., Volpe, S., Zaveri, J., Barrow, M. P., Gunn, G . B., et al. (2018). Radiotherapy dose–volume parameters predict videoﬂuoroscopy -detected dysphagia per DIGEST after IMRT for oropharyngeal cancer: Results of a pros pective registry. Radio- therapy and Oncology 128, 442–451

work page 2018

[13] [13]

Kang, L., Xiong, C., Crane, P., and Tian, L. (2013). Linear combinatio ns of biomarkers to improve diagnostic accuracy with three ordinal diagnostic categ ories. Statistics in Medicine 32, 631–643. 24 Biometrics,

work page 2013

[14] [14]

Karmakar, N. (1984). New polynomial-time algorithm for linear progr amming. Combina- torica 4, 373–395

work page 1984

[15] [15]

Kerr, C., Dura-Bernal, S., Smolinski, T., Chadderdon, G., and Wilson, D . (2018). Optimiza- tion by adaptive stochastic descent. PLOS One 13, e0192944

work page 2018

[16] [16]

Kirkpatrick, S., Gelatt, C., and Vecchi, M. (1983). Optimization by sim ulated annealing. Australian Journal of Biological Sciences 220, 671–680

work page 1983

[17] [17]

C., Teguh, D

Levendag, P. C., Teguh, D. N., Voet, P., van der Est, H., Noever, I., de Kruijf, W. J., et al. (2007). Dysphagia disorders in patients with cancer of the oropha rynx are signiﬁcantly aﬀected by the radiation therapy dose to the superior and middle co nstrictor muscle: a dose-eﬀect relationship. Radiotherapy and Oncology 85, 64–73

work page 2007

[18] [18]

and Fine, J

Li, J. and Fine, J. (2008). ROC analysis with multiple classes and multiple tests: methodology and its application in microarray studies. Biostatistics 9, 566–576

work page 2008

[19] [19]

Liu, C., Liu, A., and Halabi, S. (2011). A min–max combination of biomark ers to improve diagnostic accuracy. Statistics in Medicine 30, 2005–2014

work page 2011

[20] [20]

and Xiong, C

Luo, J. and Xiong, C. (2012). Diagtest3grp: An r package for ana lyzing diagnostic tests with three ordinal groups. J Stat Softw. 51, 1–24

work page 2012

[21] [21]

and Huang, J

Ma, S. and Huang, J. (2007). Combining multiple markers for classiﬁc ation using ROC. Biometrics 63, 751–757

work page 2007

[22] [22]

Maiti, R., Li, J., Das, P., Feng, L., Hausenloy, D., and Chakraborty, B. (2019). A distribution- free smoothed combination method of biomarkers to improve diagno stic accuracy in multi-category classiﬁcation. arxiv.org/abs/1904.10046

work page internal anchor Pith review Pith/arXiv arXiv 2019

[23] [23]

Mossman, D. (1999). Three-way ROCs. Medical Decision Making 19, 78–89

work page 1999

[24] [24]

and Yiannoutsos, C

Nakas, C. and Yiannoutsos, C. (2004). Ordered multiple-class ROC analysis with continuous measurements. Statistics in Medicine 23, 3437–3449

work page 2004

[25] [25]

and Mead, R

Nelder, J. and Mead, R. (1965). A simplex method for function minimiz ation. Computer Estimating the optimal linear combination of predictors us ing SCOR 25 Journal 7, 308–313

work page 1965

[26] [26]

Pepe, M., Cai, T., and Longton, G. (2006). Combining predictors for classiﬁcation using the area under the receiver operating characteristic curve. Biometrics 62, 221–229

work page 2006

[27] [27]

and Thompson, M

Pepe, M. and Thompson, M. (2000). Combining diagnostic test resu lts to increase accuracy. Biostatistics 1, 123–140. Scurﬁeld, B. (1996). Multiple-event forced-choice tasks in the th eory of signal detectability. Journal of Mathematical Psychology 40, 253–269

work page 2000

[28] [28]

and Liu, J

Su, J. and Liu, J. (1993). Linear combinations of multiple diagnostic m arkers. Journal of the American Statistical Association 88, 1350–1355

work page 1993

[29] [29]

G., Eisenhauer, E

Therasse, P., Arbuck, S. G., Eisenhauer, E. A., Wanders, J., Kaplan , R. S., Rubinstein, L., et al. (2000). New guidelines to evaluate the response to treatmen t in solid tumors. Journal of the National Cancer Institute 92, 205–216

work page 2000

[30] [30]

Torczon, V. (1997). On the convergence of pattern search algo rithms. SIAM Journal on Optimization 7, 1–25

work page 1997

[31] [31]

D., Setser, A., Rusch, V., Jaques, D., Budach , V., et al

Trotti, A., Colevas, A. D., Setser, A., Rusch, V., Jaques, D., Budach , V., et al. (2003). Ctcae v3.0: development of a comprehensive grading system for the adve rse eﬀects of cancer treatment. In Seminars in Radiation Oncology , volume 13, 176–181

work page 2003

[32] [32]

Yanyu, Z. (2010). ROC analysis in diagnostic medicine (phd thesis). Department of Statistics and Applied Probability, National University of Singapore

work page 2010

[33] [33]

Youden, W. (1950). Index for rating diagnostic tests. Cancer 3, 32–35

work page 1950

[34] [34]

and Li, J

Zhang, Y. and Li, J. (2011). Combining multiple markers for multi-cat egory classiﬁcation: An ROC surface approach. Australian & New Zealand Journal of Statistics 53, 63–78

work page 2011

[35] [35]

Zhou, X.-H., McClish, D., and Obuchowski, N. (2002). Statistical Methods in Diagnostic Medicine, volume 464. John Wiley and Sons. Received . . . 2020. Revised . . . 2020. Accepted . . . 2020. 26 Biometrics, . . . . . . Figure 1. Flowchart of the SCOR algorithm where s denotes the step size, φ denotes the step size threshold, and SOL(k) denotes the solutio...

work page 2002