Article
Physics, Fluids & Plasmas
Jonathan Larson, Jukka-Pekka Onnela
Summary: Mechanistic network models allow researchers to study complex systems, but it is challenging to estimate the parameters for graphs generated with these models. This paper proposes a method of treating the node sequence as an additional parameter or a missing random variable in growing network models and maximizing the resulting likelihood. The proposed framework is tested on simulated graphs and applied to protein-protein interaction networks.
Article
Computer Science, Artificial Intelligence
Yingjin Song, Daniel Beck
Summary: Most previous work in music emotion recognition assumes a single or a few song-level labels for the whole song. In this study, we propose a method to predict emotion dynamics in song lyrics without song-level supervision. Our experiments show that applying our method consistently improves the performance of sentence-level baselines without requiring any annotated songs, making it ideal for limited training data scenarios. Further analysis through case studies shows the benefits of our method while also indicating the limitations and pointing to future directions.
TRANSACTIONS OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS
(2023)
Article
Statistics & Probability
Qingguo Tang, Rohana J. Karunamuni, Boxiao Liu
Summary: In this paper, we investigate the application of robust parameter estimation and variable selection in binary regression models. We propose regularized minimum-distance estimators based on the minimum-distance approach using minimum Hellinger and minimum symmetric chi-squared distances criteria. Our study shows that these estimators are efficient and possess excellent robustness properties under the model. Through Monte Carlo studies, we analyze the small-sample and robustness properties of the proposed estimators and compare them with traditional likelihood estimators. Two real-data applications illustrate the satisfactory performance of our methods.
JOURNAL OF APPLIED STATISTICS
(2022)
Article
Engineering, Civil
I. Ben Nasr, F. Chebana
Summary: Hydrological extreme events are composed of several correlated variables, and the dependence structure between these variables needs to be considered for better risk assessment using copulas. Mixture copula is suitable for extreme events generated from different phenomena, but existing parameter estimation methods for mixture copula have drawbacks. To overcome these drawbacks, a new parameter estimation approach based on the maximum pseudo-likelihood using a metaheuristic algorithm is proposed. Simulation and real data studies show that the proposed method can estimate parameters accurately even with small sample sizes compared to existing methods.
JOURNAL OF HYDROLOGY
(2022)
Article
Economics
Fa Wang
Summary: This paper revisits the main results in Bai (2003) and Bai and Ng (2006) for generalized factor models, with slightly stronger conditions. It establishes convergence rates and asymptotic normality of the estimated factor space and loading space under mild conditions that allow for various single-index nonlinear models. Mixed models are also considered as the probability density/mass function can vary across subjects and time. For factor-augmented regressions, the limit distributions of parameter estimates, conditional mean, and forecast are derived when factors estimated from nonlinear/mixed data are used as proxies for the true factors.
JOURNAL OF ECONOMETRICS
(2022)
Article
Automation & Control Systems
Abdullah Yalcinkaya, Iklim Gedik Balay, Birdal Senoglu
Summary: This study employs genetic algorithm to obtain maximum likelihood estimates in multiple linear regression, and finds that GA performs better than traditional algorithms when the error distribution is long-tailed symmetric. The use of robust confidence intervals based on modified ML estimators as the search space in GA improves parameter estimation efficiency, as indicated by the simulation results.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2021)
Article
Geography
Chiara Ghiringhelli, Gianfranco Piras, Giuseppe Arbia, Antonietta Mira
Summary: In this paper, a recursive approach is proposed for estimating the spatial error model. The suggested methodology is compared with standard estimation procedures, and a series of Monte Carlo experiments demonstrate that the recursive approach significantly reduces computational effort while maintaining reasonable precision of the estimators. This technique proves valuable for analyzing real-time geographical data streams, especially in the era of big data. Finally, the methodology is illustrated using earthquake data.
GEOGRAPHICAL ANALYSIS
(2023)
Article
Economics
Samuele Centorrino, Maria Perez-Urdiales
Summary: We propose and study a maximum likelihood estimator for stochastic frontier models with endogeneity in cross-section data. The composite error term may be correlated with inputs and environmental variables. Our framework is a generalization of the normal half-normal stochastic frontier model with endogeneity. We derive the likelihood function in closed form and provide a computationally fast and easy-to-implement estimator. We also analyze its asymptotic properties and demonstrate its performance in finite samples through simulations and an empirical application.
JOURNAL OF ECONOMETRICS
(2023)
Article
Mathematics
Opeyo Peter Otieno, Weihu Cheng
Summary: Convergence of the maximization algorithm in logistic regression models is critical, but may fail. Bias correction methods have been conducted for maximum likelihood estimates of parameters for complete data sets and longitudinal models. Balanced data sets yield consistent estimates from conditional Logit estimators for binary response panel data models.
Article
Health Care Sciences & Services
Chew-Seng Chee, Il Do Ha, Byungtae Seo, Youngjo Lee
Summary: The paper proposes a nonparametric maximum likelihood method for a general class of nonparametric frailty models, demonstrating the effectiveness of the proposed method in improving the accuracy and robustness of regression coefficient estimation under various frailty distributions.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Operations Research & Management Science
G. Jacinto, P. A. Filipe, C. A. Braumann
Summary: We used a class of stochastic differential equations to model individual growth in a randomly fluctuating environment using cattle weight data. Maximum likelihood theory was applied to estimate the parameters, but for cattle data, it is often difficult to obtain observations at equally spaced ages or even at the same ages for different animals, leading to inaccurate maximum likelihood estimates. To improve the estimation, we introduced a weight function associated with the elapsed times between observations of each animal in the likelihood function, resulting in the weighted maximum likelihood method. Comparing the results from both methods in different data structures, we found that the weighted maximum likelihood method performs better when there are few observations at older ages and the observation instants are unequally spaced. For unequally spaced observations, a bootstrap estimation method was also applied and proved to be a more precise alternative, except when the available data only includes young animals.
Article
Multidisciplinary Sciences
Chenangnon Frederic Tovissode, Aliou Diop, Romain Glele Kakai
Summary: The study introduced a GLMM method based on skew generalized t distributions that outperforms traditional approaches in estimating population parameters and predicting random effects; Utilizing the Expectation-Maximization algorithm for model fitting and parameter expansion; In an application with respiratory infection data, the superiority of the method was demonstrated as the most adequate model.
Article
Multidisciplinary Sciences
Jie You, Zhaoxuan Li, Junli Du
Summary: This paper proposes a new iterative method of EM initialization (MRIPEM) to address the sensitivity and local optimum problems in Gaussian mixture model parameter estimation. The mean vector and covariance matrix of the sample are used as initial values and continuously updated through clustering based on the maximum Mahalanobis distance. Experimental results show that the MRIPEM algorithm is comparable to other popular initialization methods in relatively high dimensions and overlaps, and significantly better in low dimensions and overlaps.
Article
Computer Science, Interdisciplinary Applications
Jinhyeun Kim, Christopher Luettgen, Kamran Paynabar, Fani Boukouvala
Summary: In Gaussian Process Regression (GPR), embedding physics-based knowledge through penalization of the marginal likelihood function can improve the prediction performance, reduce violation of known physics, and mitigate overfitting problems. This paper presents three case studies where physics-based knowledge is available in the form of linear Partial Differential Equations (PDEs), and shows that the new hyperparameter set obtained from the augmented marginal likelihood function leads to consistent optimal hyperparameters and quality GPR fit, despite the challenge of unknown initial or boundary conditions.
COMPUTERS & CHEMICAL ENGINEERING
(2023)
Article
Mathematics
Josmar Mazucheli, Bruna Alves, Mustafa C. Korkmaz, Victor Leiva
Summary: The Vasicek distribution is a two-parameter probability model that plays an important role in statistical applications, particularly in finance. This paper proposes two Vasicek regression models for analyzing data on the unit interval, one using a quantile parameterization and the other using the original parameterization. Monte Carlo simulations are conducted to evaluate the statistical properties of the estimators, and an R package is developed to provide the results of the investigation. Applications with real data sets demonstrate the practical usage of the Vasicek quantile and mean regressions as alternatives to other well-known models.
Article
Environmental Sciences
Ram B. Jain
Summary: This study analyzed data from NHANES for US adults aged 20 years and above to estimate cotinine levels among different groups of smokers. Results showed that males had higher cotinine levels in certain groups than females, and non-Hispanic black smokers had higher levels compared to non-Hispanic whites. Additionally, estimations were made for self-reported nonsmokers classified as smokers and smokers with missing self-reported data on tobacco product use.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: Analysis of data from the National Health and Nutrition Examination Survey for US adults aged 20 years and older between 2005 and 2016 showed variations in concentrations of arsenobetaine, monomethylarsonic acid, dimethylarsenic acid, and total arsenic in urine across different stages of renal function. Differences were observed in concentrations based on gender and racial/ethnic groups, with levels decreasing over the survey years.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: The study found that glomerular hyperfiltration has a significant impact on the concentrations of perfluoroalkyl acids, with hyperfiltrators generally having lower adjusted geometric means compared to normal filtrators across various disease groups. Male-female differences in adjusted geometric means were usually narrower for normal filtrators, and the disease group with hypertension only had the highest adjusted geometric means for every perfluoroalkyl acid. Additionally, among hyperfiltrators, the group with anemia only had the lowest adjusted geometric means for every perfluoroalkyl acid compared to other disease groups.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain, Alan Ducatman
Summary: This study found a positive association between PFAS exposure and Apo B, especially in non-diabetic individuals not taking lipid lowering medications. Diabetic individuals showed a greater impact of lipid lowering medications on Apo B compared to non-diabetic populations. Further research is needed to replicate these findings in other populations and to explore mechanistic studies.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: The study revealed a negative association between blood cadmium concentrations and certain PFAAs, a positive association between blood lead concentrations and most PFAAs, and a positive association between blood total mercury concentrations and all PFAAs.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: This study found higher cotinine concentrations in males than females for children, adolescent smokers, and nonsmoker adults. Non-Hispanic Blacks had lower concentrations of both cotinine and hydroxycotinine than non-Hispanic Whites for adult smokers, and the ratio of hydroxycotinine concentrations for those exposed to ETS at home compared to those not exposed was different for nonsmoker adults and adult smokers.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: The study evaluated the variations in concentrations of selected monohydroxy polycyclic aromatic hydrocarbons in urine across different stages of glomerular function, finding differences in how the kidneys process PAH metabolites among smokers and nonsmokers.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: The study analyzed data from 639 US children aged 3-11 years and found significant associations between PFAAs and blood lead and mercury levels, suggesting potential co-exposure to PFAAs and lead/mercury may lead to more severe neurodevelopmental deficits. Additional research is needed to further investigate the additive/synergistic neurodevelopmental deficits associated with co-exposures to PFAAs and lead/mercury.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2021)
Article
Environmental Sciences
Ram B. Jain
Summary: Exposure to cadmium and lead can cause oxidative stress and result in kidney and cardiovascular diseases. The antiaging protein klotho acts as an antioxidant. This study found that the concentrations of klotho were affected by cadmium and lead exposure, with an observed decrease in klotho concentrations during kidney dysfunction, particularly with blood cadmium concentrations.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2022)
Article
Environmental Sciences
Ram B. Jain, Alan Ducatman
Summary: This study examined the associations between blood manganese and selenium with serum concentrations of perfluoroalkyl substances (PFAS). Among adults, blood manganese concentrations were inversely associated with serum concentrations of several PFAS, while blood selenium concentrations were positively associated with PFAS in adults and adolescents.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2022)
Article
Environmental Sciences
Ram B. Jain
Summary: This study analyzed data from the US National Health and Nutrition Examination Survey to investigate gender and racial/ethnic differences in PFHpS concentrations among US adults. The study found that males had significantly higher concentrations of PFHpS than females, and PFHpS concentrations followed an inverted U-shaped curve across different stages of kidney function.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2023)
Article
Environmental Sciences
Ram B. Jain, Alan Ducatman
Summary: This study used nationally representative data for US from 2003 to 2018 to analyze the gender-based differences in serum concentrations of PFAS. The results showed that females had lower serum PFAS levels than males at certain age ranges, and the differences were maximized at different ages for different compounds. The findings suggest the importance of separate analyses of male and female data, as well as stratified analysis for different time periods in females. The study also provides support for further research on the influences of gender differences in serum PFAS.
SCIENCE OF THE TOTAL ENVIRONMENT
(2022)
Article
Environmental Sciences
Ram B. Jain
Summary: Data from the National Health and Nutrition Examination Survey were analyzed to explore the relationships between certain fluorinated carbon compounds and urinary concentrations of arsenic. The results showed positive associations between PFNA and all four arsenic variables, with statistical significance observed only for IAS. PFDA and PFUnDA were positively associated with urinary arsenic, while Me-PFOSA with PFAS showed inverse associations, with significance observed only for UDMA. PFOA, PFHxS, and PFOS generally exhibited negative associations with arsenic, but without statistical significance. Further investigation is needed to understand the impact of co-exposure to PFAS and arsenic on health, with fluorinated carbon chain length potentially playing a role in defining these associations.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2023)
Article
Environmental Sciences
Ram B. Jain
Summary: The associations between urinary concentrations of oxidant polycyclic aromatic hydrocarbon (PAH) metabolites and serum concentrations of anti-oxidant alpha-klotho were examined in US adults aged 40-79 years. The study found that increased PAH metabolite concentrations were associated with decreased alpha-klotho concentrations in individuals with normal or near normal kidney function. However, the associations were not significant in individuals with albuminuria. These findings suggest that exposure to PAH may lead to reduced alpha-klotho concentrations.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2023)
Article
Environmental Sciences
Ram B. Jain
Summary: This study estimated the associations between urinary concentrations of oxidant polycyclic aromatic hydrocarbon (PAH) metabolites and serum concentrations of anti-oxidant alpha-klotho in US adults aged 40-79 years. The results showed that increased levels of certain PAH metabolites were associated with decreased serum alpha-klotho concentrations, especially in individuals without albuminuria and normal or near normal kidney function.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2022)