Paul Vos
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Paul Vos has a total of 109 co-authors. Here are the top co-authors by number of shared publications:
- Nam Phong Nguyen
Department of Radiation Oncology, Howard University, Washington, DC 20060, USA
10 shared publications - Ulf Karlsson
Department of Radiation Oncology, International Geriatric Radiotherapy Group, Washington, DC 20001, USA
8 shared publications - Sabah Sallah
Division of Hematology/Oncology, University of Tennessee Health Science Center, Memphis 38103, USA
6 shared publications Cheryl Frank
5 shared publicationsLy M Nguyen
5 shared publications- Suresh Dutta
11Medicine and Radiation Oncology PA, San Antonio, TX, USA
5 shared publications
Latest Publications
- Open Access
- 7 September 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (17) , 3237
Rethinking Mean Square Error: Information, Generalized Estimation, and the James–Stein Paradox
The James–Stein estimator’s dominance over maximum likelihood in mean square error has been called a paradox because maximum likelihood is known to be superior in many other respects. One response, due to Efron, is to question maximum likelihood. Another is to question MSE. We pursue the second and compare MSE with -information as criteria for assessing estimators. The comparison rests on two distinctions: between point estimators and generalized estimators—functions of the sample and parameter jointly, with the score as archetype—as inferential objects, and between pointwise and family-aware assessment criteria. An elementary lemma shows that no pointwise criterion, MSE or any other risk built from a loss function, admits a uniformly optimal estimator; -information, which is family-aware and parameter-invariant, is uniformly maximized by the score. A point estimator is assessed through the generalized estimators it induces, and under the score map, its -efficiency is the fraction of Fisher information the statistic retains, placing the criterion in Fisher’s information-loss tradition. A point estimator in fact induces two such objects, one through its mean function and one through its own marginal score; the second is never less informative than the first, with equality exactly when the statistic is the natural statistic of its own marginal exponential family. On unbiased estimators, -efficiency coincides with variance-based efficiency. Returning to James–Stein, the paradox dissolves: maximum likelihood is fully efficient because it is sufficient, while the James–Stein statistic is exactly two-to-one in the sample, and the information it destroys—computed exactly—is concentrated precisely where its MSE advantage is greatest, so that what the shrinkage gains for a point estimate, it withholds from any inference built on it. MSE retains its proper domain under genuine squared-error loss.- 7 September 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (17) , 3237
Open AccessRethinking Mean Square Error: Information, Generalized Estimation, and the James–Stein Paradox
The James–Stein estimator’s dominance over maximum likelihood in mean square error has been called a paradox because maximum likelihood is known to be superior in many other respects. One response, due to Efron, is to question maximum likelihood. Another is to question MSE. We pursue the second and compare MSE with -information as criteria for assessing estimators. The comparison rests on two distinctions: between point estimators and generalized estimators—functions of the sample and parameter jointly, with the score as archetype—as inferential objects, and between pointwise and family-aware assessment criteria. An elementary lemma shows that no pointwise criterion, MSE or any other risk built from a loss function, admits a uniformly optimal estimator; -information, which is family-aware and parameter-invariant, is uniformly maximized by the score. A point estimator is assessed through the generalized estimators it induces, and under the score map, its -efficiency is the fraction of Fisher information the statistic retains, placing the criterion in Fisher’s information-loss tradition. A point estimator in fact induces two such objects, one through its mean function and one through its own marginal score; the second is never less informative than the first, with equality exactly when the statistic is the natural statistic of its own marginal exponential family. On unbiased estimators, -efficiency coincides with variance-based efficiency. Returning to James–Stein, the paradox dissolves: maximum likelihood is fully efficient because it is sufficient, while the James–Stein statistic is exactly two-to-one in the sample, and the information it destroys—computed exactly—is concentrated precisely where its MSE advantage is greatest, so that what the shrinkage gains for a point estimate, it withholds from any inference built on it. MSE retains its proper domain under genuine squared-error loss. - Open Access
- 3 August 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (15) , 2752
Strong Likelihood Principle: Strengthening a Principle or Misunderstanding the Likelihood Function
The strong likelihood principle (SLP) is conventionally derived from the sufficiency principle and a conditionality principle in an argument due to Birnbaum, and much of the literature contests whether the derivation is sound. We take a different approach. We ask what the SLP says when its terms are read carefully, and argue that the principle as ordinarily stated reflects a confusion about the domain of the likelihood function. The likelihood is naturally defined as a function on a family of distributions M, not on a parameter space, and once it is so defined the SLP collapses into its weak counterpart, the weak likelihood principle. The diagnosis is illustrated by analogy with monetary value, developed concretely through a comparison of the binomial and negative binomial families that share a parameter, and connected to the geometric structure of M through the Fisher information metric. The same standardization emerges from a statistical argument about comparing measurements across populations and from a geometric argument about manifold distance; this convergence supplies the positive content of the weak likelihood principle.
- 3 August 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (15) , 2752
Open AccessStrong Likelihood Principle: Strengthening a Principle or Misunderstanding the Likelihood Function
The strong likelihood principle (SLP) is conventionally derived from the sufficiency principle and a conditionality principle in an argument due to Birnbaum, and much of the literature contests whether the derivation is sound. We take a different approach. We ask what the SLP says when its terms are read carefully, and argue that the principle as ordinarily stated reflects a confusion about the domain of the likelihood function. The likelihood is naturally defined as a function on a family of distributions M, not on a parameter space, and once it is so defined the SLP collapses into its weak counterpart, the weak likelihood principle. The diagnosis is illustrated by analogy with monetary value, developed concretely through a comparison of the binomial and negative binomial families that share a parameter, and connected to the geometric structure of M through the Fisher information metric. The same standardization emerges from a statistical argument about comparing measurements across populations and from a geometric argument about manifold distance; this convergence supplies the positive content of the weak likelihood principle.
- Open Access
- 23 December 2025
- Published by MDPI AG in Mathematics
- Vol. 14 (1) , 45
Confidence Intervals for Two Proportions—A Generalized Estimation and Information Assessment
This paper develops profile score confidence intervals (i.e., z-standard intervals) by inverting orthogonalized score estimators for comparing two independent binomial proportions in rate difference, rate ratio, or odds ratio and utilizes a generalized estimation framework of Vos and Wu (Inf. Geom. 8: 99–123, 2025) to evaluate different confidence interval methods. The orthogonalized score estimators dominate other generalized estimators in -efficiency for distinguishing parameter values around the truth, so the z-standard intervals become more efficient and acquire coverage closer to the nominal level than other types of confidence intervals. In addition, the degrees of freedom and small-sample corrections (applied to the profile nuisance parameter estimates) are expected to improve the coverage of the z-standard intervals and help maintain them above the nominal level. Computational algorithms are developed to find the z-standard intervals using R’s polyroot function. Numerical studies are conducted to compare both coverage and endpoints of different types of confidence intervals.- 23 December 2025
- Published by MDPI AG in Mathematics
- Vol. 14 (1) , 45
Open AccessConfidence Intervals for Two Proportions—A Generalized Estimation and Information Assessment
This paper develops profile score confidence intervals (i.e., z-standard intervals) by inverting orthogonalized score estimators for comparing two independent binomial proportions in rate difference, rate ratio, or odds ratio and utilizes a generalized estimation framework of Vos and Wu (Inf. Geom. 8: 99–123, 2025) to evaluate different confidence interval methods. The orthogonalized score estimators dominate other generalized estimators in -efficiency for distinguishing parameter values around the truth, so the z-standard intervals become more efficient and acquire coverage closer to the nominal level than other types of confidence intervals. In addition, the degrees of freedom and small-sample corrections (applied to the profile nuisance parameter estimates) are expected to improve the coverage of the z-standard intervals and help maintain them above the nominal level. Computational algorithms are developed to find the z-standard intervals using R’s polyroot function. Numerical studies are conducted to compare both coverage and endpoints of different types of confidence intervals. - Open Access
Geometry of Statistical Manifolds
A statistical manifold M can be defined as a Riemannian manifold each of whose points is a probability distribution on the same support. In fact, statistical manifolds possess a richer geometric structure beyond the Fisher information metric defined on the tangent bundle . Recognizing that points in M are distributions and not just generic points in a manifold, can be extended to a Hilbert bundle . This extension proves fundamental when we generalize the classical notion of a point estimate—a single point in M—to a function on M that characterizes the relationship between observed data and each distribution in M. The log likelihood and score functions are important examples of generalized estimators. In terms of a parameterization , is a distribution on while its generalization as an estimate is a function over that indicates inconsistency between the model and data. As an estimator, is a distribution of functions. Geometric properties of these functions describe statistical properties of . In particular, the expected slopes of are used to define , the -information of . The Fisher information I is an upper bound for the -information: for all g, . We demonstrate the utility of this geometric perspective using the two-sample problem.Open AccessGeometry of Statistical Manifolds
A statistical manifold M can be defined as a Riemannian manifold each of whose points is a probability distribution on the same support. In fact, statistical manifolds possess a richer geometric structure beyond the Fisher information metric defined on the tangent bundle . Recognizing that points in M are distributions and not just generic points in a manifold, can be extended to a Hilbert bundle . This extension proves fundamental when we generalize the classical notion of a point estimate—a single point in M—to a function on M that characterizes the relationship between observed data and each distribution in M. The log likelihood and score functions are important examples of generalized estimators. In terms of a parameterization , is a distribution on while its generalization as an estimate is a function over that indicates inconsistency between the model and data. As an estimator, is a distribution of functions. Geometric properties of these functions describe statistical properties of . In particular, the expected slopes of are used to define , the -information of . The Fisher information I is an upper bound for the -information: for all g, . We demonstrate the utility of this geometric perspective using the two-sample problem. - Open Access
- 19 May 2025
- Published by Cambridge University Press (CUP) in Antimicrobial Stewardship & Healthcare Epidemiology
- Vol. 5 (1) , e117
Comparison of intravenous vs intravenous with step-down to oral antibiotic treatment course for Streptococcus and Enterococcus bloodstream infections
Objective: To compare clinical failure of intravenous vs intravenous with oral step-down antibiotic treatment for Streptococcus and Enterococcus bloodstream infection. Design and setting: Multicenter, retrospective, cohort study at one academic medical center and eight community hospitals. Patients: Hospitalized adult patients with blood cultures positive for Streptococcus or Enterococcus were included. Patients were excluded if they had complicated infection, had polymicrobial bacteremia, received less than 5 days of therapy, or died before completing therapy. Methods: Patients who completed intravenous therapy were compared with patients who transitioned to oral therapy after 3 to 7 days. The primary endpoint was clinical failure, defined as 90-day all-cause mortality or recurrent bacteremia. The primary analysis excluded patients with unknown outcomes, and the sensitivity analysis treated them as failures. Results: 429 patients were included (intravenous group: n = 225; oral step-down group; n = 204). The intravenous group had more comorbidities and vasopressor use. The intravenous group had a higher risk of clinical failure in the primary analysis (17.5% vs. 8.8%; adjusted OR 2.14 [95% CI, 1.09–4.2]; p = 0.03) while the sensitivity analysis found no difference in clinical failure (adjusted OR 1.1 [95% CI, 0.69–1.74], p = 0.69). The oral step-down group had a mean length of stay of 9.2 days shorter than the intravenous group ([95% CI, 7.5–11.0]; p<0.001). Conclusion: Oral step-down therapy was not associated with an increased risk of clinical failure compared to a full course of intravenous therapy for uncomplicated Streptococcus and Enterococcus bloodstream infections. Patients with more comorbidities or who required vasopressors were less likely to be switched to oral therapy.
- 19 May 2025
- Published by Cambridge University Press (CUP) in Antimicrobial Stewardship & Healthcare Epidemiology
- Vol. 5 (1) , e117
Open AccessComparison of intravenous vs intravenous with step-down to oral antibiotic treatment course for Streptococcus and Enterococcus bloodstream infections
Objective: To compare clinical failure of intravenous vs intravenous with oral step-down antibiotic treatment for Streptococcus and Enterococcus bloodstream infection. Design and setting: Multicenter, retrospective, cohort study at one academic medical center and eight community hospitals. Patients: Hospitalized adult patients with blood cultures positive for Streptococcus or Enterococcus were included. Patients were excluded if they had complicated infection, had polymicrobial bacteremia, received less than 5 days of therapy, or died before completing therapy. Methods: Patients who completed intravenous therapy were compared with patients who transitioned to oral therapy after 3 to 7 days. The primary endpoint was clinical failure, defined as 90-day all-cause mortality or recurrent bacteremia. The primary analysis excluded patients with unknown outcomes, and the sensitivity analysis treated them as failures. Results: 429 patients were included (intravenous group: n = 225; oral step-down group; n = 204). The intravenous group had more comorbidities and vasopressor use. The intravenous group had a higher risk of clinical failure in the primary analysis (17.5% vs. 8.8%; adjusted OR 2.14 [95% CI, 1.09–4.2]; p = 0.03) while the sensitivity analysis found no difference in clinical failure (adjusted OR 1.1 [95% CI, 0.69–1.74], p = 0.69). The oral step-down group had a mean length of stay of 9.2 days shorter than the intravenous group ([95% CI, 7.5–11.0]; p<0.001). Conclusion: Oral step-down therapy was not associated with an increased risk of clinical failure compared to a full course of intravenous therapy for uncomplicated Streptococcus and Enterococcus bloodstream infections. Patients with more comorbidities or who required vasopressors were less likely to be switched to oral therapy.
- Open Access
- 26 February 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (5) , 775
Random Variables Aren’t Random
This paper examines the foundational concept of random variables in probability theory and statistical inference, demonstrating that their mathematical definition requires no reference to randomization or hypothetical repeated sampling. We show how measure-theoretic probability provides a framework for modeling populations through distributions, leading to three key contributions. First, we establish that random variables, properly understood as measurable functions, can be fully characterized without appealing to infinite hypothetical samples. Second, we demonstrate how this perspective enables statistical inference through logical rather than probabilistic reasoning, extending the reductio ad absurdum argument from deductive to inductive inference. Third, we show how this framework naturally leads to an information-based assessment of statistical procedures, replacing traditional inference metrics that emphasize bias and variance with information-based approaches that describe the families of distributions used in parametric inference better. This reformulation addresses long-standing debates in statistical inference while providing a more coherent theoretical foundation. Our approach offers an alternative to traditional frequentist inference that maintains mathematical rigor while avoiding the philosophical complications inherent in repeated sampling interpretations.
- 26 February 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (5) , 775
Open AccessRandom Variables Aren’t Random
This paper examines the foundational concept of random variables in probability theory and statistical inference, demonstrating that their mathematical definition requires no reference to randomization or hypothetical repeated sampling. We show how measure-theoretic probability provides a framework for modeling populations through distributions, leading to three key contributions. First, we establish that random variables, properly understood as measurable functions, can be fully characterized without appealing to infinite hypothetical samples. Second, we demonstrate how this perspective enables statistical inference through logical rather than probabilistic reasoning, extending the reductio ad absurdum argument from deductive to inductive inference. Third, we show how this framework naturally leads to an information-based assessment of statistical procedures, replacing traditional inference metrics that emphasize bias and variance with information-based approaches that describe the families of distributions used in parametric inference better. This reformulation addresses long-standing debates in statistical inference while providing a more coherent theoretical foundation. Our approach offers an alternative to traditional frequentist inference that maintains mathematical rigor while avoiding the philosophical complications inherent in repeated sampling interpretations.
- Open Access
- 22 February 2025
- Published by Springer Nature in Information Geometry
- Vol. 8 (1) , 99-123
Generalized estimation and information
This paper extends the idea of a generalized estimator for a scalar parameter (Vos in Inform Geom 7:151–170, 2022) to multi-dimensional parameters both with and without nuisance parameters. The title reflects the fact that generalized estimators provide more than simply another method to find point estimators, and that the methods to assess generalized estimators differ from those for point estimators. By generalized estimation we mean the use of generalized estimators together with an extended definition of information to assess their inferential properties. We show that Fisher information provides an upper bound for the information utilized by an estimator and that the score attains this bound. This optimality result provides theoretical justification for likelihood-based inference, effectively narrowing the search for optimal estimators when the score is computationally feasible.
- 22 February 2025
- Published by Springer Nature in Information Geometry
- Vol. 8 (1) , 99-123
Open AccessGeneralized estimation and information
This paper extends the idea of a generalized estimator for a scalar parameter (Vos in Inform Geom 7:151–170, 2022) to multi-dimensional parameters both with and without nuisance parameters. The title reflects the fact that generalized estimators provide more than simply another method to find point estimators, and that the methods to assess generalized estimators differ from those for point estimators. By generalized estimation we mean the use of generalized estimators together with an extended definition of information to assess their inferential properties. We show that Fisher information provides an upper bound for the information utilized by an estimator and that the score attains this bound. This optimality result provides theoretical justification for likelihood-based inference, effectively narrowing the search for optimal estimators when the score is computationally feasible.
- Open Access
- 13 May 2023
- Published by Wiley in Scandinavian Journal of Statistics
- Vol. 50 (3) , 920-922
Comments on Divergence vs. Decision P‐values
- 13 May 2023
- Published by Wiley in Scandinavian Journal of Statistics
- Vol. 50 (3) , 920-922
Open AccessComments on Divergence vs. Decision P‐values
- 4 December 2022
- Published by Springer Nature in Information Geometry
- Vol. 7 (S1) , 151-170
Generalized estimators, slope, efficiency, and fisher information bounds
- 4 December 2022
- Published by Springer Nature in Information Geometry
- Vol. 7 (S1) , 151-170
Generalized estimators, slope, efficiency, and fisher information bounds
- 1 October 2022
- Published by IOP Publishing in Medical Engineering & Physics
- Vol. 109 (1) , 103899
Electromyography as a surrogate for estimating metabolic energy expenditure during locomotion
Minimizing metabolic energy expenditure (MEE) plays an important role in increasing mobility in people with locomotor disabilities, as movements that require high energy lead to less activity. Rehabilitation programs and devices use MEE to determine how effective they are, but using indirect calorimetry is limiting due to time delays and non-real-world conditions. Electromyography (EMG) offers insight into how muscles activate; thus, the purpose of this study was to develop a real-time MEE feedback system through the utilization of EMG signals. Participants completed five walking conditions at different stride frequencies (preferred, +/- 15%, +/- 30%), while breath-by-breath gas exchange, ground reaction forces and EMG signals were collected. The live EMG signal was numerically integrated and separated into strides, then scaled by a cost of force (COF) coefficient. MEE had the expected quadratic relationship seen in previous literature (R2 = 0.967), along with COF data (R2 = 0.701). The EMG method stabilized between 75.1% - 133.1%, which is not within a close range (90% - 110%) of MEE; thus, future studies must investigate other mathematical methods. Our results indicate a qualitative association between MEE and EMG activity, which could be used to increase mobility and quality of life for populations with disability.
- 1 October 2022
- Published by IOP Publishing in Medical Engineering & Physics
- Vol. 109 (1) , 103899
Electromyography as a surrogate for estimating metabolic energy expenditure during locomotion
Minimizing metabolic energy expenditure (MEE) plays an important role in increasing mobility in people with locomotor disabilities, as movements that require high energy lead to less activity. Rehabilitation programs and devices use MEE to determine how effective they are, but using indirect calorimetry is limiting due to time delays and non-real-world conditions. Electromyography (EMG) offers insight into how muscles activate; thus, the purpose of this study was to develop a real-time MEE feedback system through the utilization of EMG signals. Participants completed five walking conditions at different stride frequencies (preferred, +/- 15%, +/- 30%), while breath-by-breath gas exchange, ground reaction forces and EMG signals were collected. The live EMG signal was numerically integrated and separated into strides, then scaled by a cost of force (COF) coefficient. MEE had the expected quadratic relationship seen in previous literature (R2 = 0.967), along with COF data (R2 = 0.701). The EMG method stabilized between 75.1% - 133.1%, which is not within a close range (90% - 110%) of MEE; thus, future studies must investigate other mathematical methods. Our results indicate a qualitative association between MEE and EMG activity, which could be used to increase mobility and quality of life for populations with disability.