Model Comparison Criteria Research Articles

Comparison of Nonlinear Functions to Define the Growth in Intensive Feedlot System with XGBoost Algorithm

The aim of this study was to define the growth by using nonlinear functions in intensive feedlot system with XGBoost algorithm. To achieve this aim, five nonlinear functions were implemented. To implementation of the study, Brown Swiss (n=41) and Simental (n=95) breed were used. Each nonlinear functions were examined for each breed. According to the results of the nonlinear functions, logistic model was the best prediction model for defining the growth of each breed. In this study, the parameters in the best prediction model were calculated individually and the relationship of these parameters with body weight was evaluated with the XGBoost algorithm. Model comparison criteria such as standard deviation ratio (SDratio), Pearson’s correlation coefficient (PC), determination of coefficient (R2) and Akaike’s information criteria (AIC) were used to evaluate the XGBoost algorithm. In conclusion, the XGBoost algorithm can be an effective and optional approach that allows breeders to estimate live weight from growth parameters. This algorithm can operate on large data sets with high accuracy and speed, leading to significant improvements in agricultural productivity and animal health management. XGBoost enables more accurate predictions by analyzing the effects of various characteristics (e.g., nutritional level, breed, age). Therefore, this method can be used to determine critical parameters such as body weight in animal breeding practices, serving as a powerful support tool for operational decisions.

Software reliability prediction: A machine learning and approximation Bayesian inference approach

AbstractReliability growth models are commonly categorized into two primary groups: parametric and non‐parametric models. Parametric models, known as Software Reliability Growth Models (SRGM) rely on a set of hypotheses that can potentially affect the accuracy of model predictions, while non‐parametric models (such as neural networks) can predict the model solely based on training data without any assumptions regarding the model itself. In this paper, we propose several methods to enhance prediction accuracy in software reliability context. More specifically, we, on one hand, introduce two gradient‐based techniques for estimating parameters of classical SRGMs. On the other, we propose methods involving LSTM Encoder–Decoder and Bayesian approximation within Langevin Gradient and Variational inference neural networks. To evaluate our proposed models' performance, we compare them with various neural network‐based software reliability models using three real‐world software failure datasets and utilizing the Mean Square Error (MSE) as a model comparison criterion. The experimental results indicate that our proposed non‐parametric models outperform most classical parametric and non‐parametric models.

Post-fire growth of Pinus halepensis: Shifts in the mode of competition along a precipitation gradient

Competition can intensify the struggle for resources among plants, affecting forest growth and dynamics. The intensity and mode of competition – asymmetric vs. symmetric – can change along environmental gradients and with time, impacting the response of plants to their environment, but this is seldom explicitly considered in management plans. In this study, we aim to (i) disentangle the main environmental and tree-related factors that affect post-fire Pinus halepensis sapling growth; (ii) determine which mode of competition characterizes the species post-fire regeneration; and (iii) elucidate if the mode of competition changes with time, and along climatic gradients. We sampled 148 P. halepensis saplings located in 15 sites affected by wildfires between 1987 and 2013 in Catalonia (NE Spain). We measured their radial growth at the base, and we identified their competitive environment by locating and measuring all the trees in a 3-meter-radius from our target saplings. We modelled the effect of tree size, age, climate, and competitive environment on the post-fire regeneration growth following the neighborhood theory of forest dynamics and using model comparison and information criteria to address our research questions. We used a modification from the Hegyi index to determine the prevalence of symmetric vs. asymmetric competition, and we assessed the support for models in which competition was allowed to vary with age and precipitation. The best model showed that competition, precipitation, age, and target size influenced the radial growth of P. halepensis post-fire regeneration (Akaike weight 0.66, R2 = 0.52). We found evidence that P. halepensis mode of competition is generally asymmetric but can change along a precipitation gradient: it shifts from asymmetric in drier sites to almost fully symmetrical as precipitation increases. Our findings have implications for the management of post-fire P. halepensis stands, enabling the adaptation of silvicultural prescriptions to the climatic environment. In a context of increasing water scarcity, adaptive silviculture is fundamental to foster the resilience of pinewoods to future climatic and disturbance regimes.

Tekrarlanan ölçümlerde farklı kovaryans yapılarının Bayes yöntemi ile modellenmesi

Objective: The objective of this study was to obtain solutions by modeling different covariance structures with Bayesian analysis methods in repeated measurement and to show its applicability to data in animal science. 
 Materials and Methods: This article focused on the analysis of the body weight data of 4154 weaned 8-month-old lambs. Repeated measurement analyses based on the mixed effect model were evaluated with Bayesian methods. Models were created for 12 different covariance structures. As the model comparison criterion, Deviation Information Criteria based on the relationship between the fit of the data to the model and the complexity of the model were used. 
 Result: Among 12 different covariance structures, the unstructured covariance structure was determined as a suitable structure for the data of this study.
 Conclusions: It was concluded that various variance-covariance structures, such as body weight, can be easily modeled in repeated measurement data. Instead of PROC MCMC methods that require complex and computational difficulties and profound coding knowledge, it was presented a relatively user-friendly and fast procedure with its theoretical structure and demonstrated its feasibility. As a result of the literature review, this is the first study in which Bayesian methods solved a wide variety of variance-covariance structure models.

An explainable artificial intelligence framework for risk prediction of COPD in smokers

BackgroundSince the inconspicuous nature of early signs associated with Chronic Obstructive Pulmonary Disease (COPD), individuals often remain unidentified, leading to suboptimal opportunities for timely prevention and treatment. The purpose of this study was to create an explainable artificial intelligence framework combining data preprocessing methods, machine learning methods, and model interpretability methods to identify people at high risk of COPD in the smoking population and to provide a reasonable interpretation of model predictions.MethodsThe data comprised questionnaire information, physical examination data and results of pulmonary function tests before and after bronchodilatation. First, the factorial analysis for mixed data (FAMD), Boruta and NRSBoundary-SMOTE resampling methods were used to solve the missing data, high dimensionality and category imbalance problems. Then, seven classification models (CatBoost, NGBoost, XGBoost, LightGBM, random forest, SVM and logistic regression) were applied to model the risk level, and the best machine learning (ML) model’s decisions were explained using the Shapley additive explanations (SHAP) method and partial dependence plot (PDP).ResultsIn the smoking population, age and 14 other variables were significant factors for predicting COPD. The CatBoost, random forest, and logistic regression models performed reasonably well in unbalanced datasets. CatBoost with NRSBoundary-SMOTE had the best classification performance in balanced datasets when composite indicators (the AUC, F1-score, and G-mean) were used as model comparison criteria. Age, COPD Assessment Test (CAT) score, gross annual income, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), anhelation, respiratory disease, central obesity, use of polluting fuel for household heating, region, use of polluting fuel for household cooking, and wheezing were important factors for predicting COPD in the smoking population.ConclusionThis study combined feature screening methods, unbalanced data processing methods, and advanced machine learning methods to enable early identification of COPD risk groups in the smoking population. COPD risk factors in the smoking population were identified using SHAP and PDP, with the goal of providing theoretical support for targeted screening strategies and smoking population self-management strategies.

Distance Educaiton Satisfaction in Higher Education Students During Covid-19 Pandemic: A Survey

The primary aim of the study was to investigate the relationship between the demographic and socioeconomic conditions of students and their satisfaction regarding distance education in Turkey. As data collection tool, a questionnaire form was created and used in order to assess the opinions of university students about distance education offered due to pandemic process. Four different models were used to investigate the significance of the relationships between the variables. Ordered logistic regression, ordered probit regression, generalized ordered logistic regression, and generalized ordered probit regression were used to determine the factors associated with the individual satisfaction levels of distance education. According to the model comparison criteria, generalized ordered logistic regression produced the best model. Further results generated by this model indicated that gender, status, age, schoolroom, marital status, location at the university, level of computer use, father’s and mother’s education level, and the university location variables had an effect on satisfaction regarding distance education. The present study determined the relationship between several factors regarding the university students, including marital status, age, income level, the region of the university, and their level of computer use, and their satisfaction level with regards to the distance education. Appropriate educational techniques should be emphasized by considering the variables that this study reveals and that have a significant relationship between them in terms of students' satisfaction with the education they receive.

Modeling and maximum likelihood based inference ofinterval-censored data with unknown upper limits andtime-dependent covariates.

Due to the nature of study design or other reasons, the upper limits of the interval-censored data with multiple visits are unknown. A naïve approach is to treat the last observed time as the exact event time, which may induce biased estimators of the model parameters. In this paper, we first develop a Cox model with time-dependent covariates for the event time and a proportional hazards model with frailty for the gap time. We then construct the upper limits using the latent gap times to resolve the issue of interval-censored event time data with unknown upper limits. A data-augmentation technique and a Monte Carlo EM (MCEM) algorithm are developed to facilitate computation. Theoretical properties of the computational algorithm are also investigated. Additionally, new model comparison criteria are developed to assess the fit of the gap time data as well as the fit of the event time data conditional on the gap time data. Our proposed method compares favorably with competing methods in both simulation study and real data analysis.

Machine learning-enabled risk prediction of chronic obstructive pulmonary disease with unbalanced data

Background and objectiveSince the early symptoms of chronic obstructive pulmonary disease (COPD) are not obvious, patients are not easily identified, causing improper time for prevention and treatment. In present study, machine learning (ML) methods were employed to construct a risk prediction model for COPD to improve its prediction efficiency. MethodsWe collected data from a sample of 5807 cases with a complete COPD diagnosis from the 2019 COPD Surveillance Program in Shanxi Province and extracted 34 potentially relevant variables from the dataset. Firstly, we used feature selection methods (i.e., Generalized elastic net, Lasso and Adaptive lasso) to select ten variables. Afterwards, we employed supervised classifiers for class imbalanced data by combining the cost-sensitive learning and SMOTE resampling methods with the ML methods (Logistic Regression, SVM, Random Forest, XGBoost, LightGBM, NGBoost and Stacking), respectively. Last, we assessed their performance. ResultsThe cough frequently at age 14 and before and other 9 variables are significant parameters for COPD. The Stacking heterogeneous ensemble model showed relatively good performance in the unbalanced datasets. The Logistic Regression with class weighting enjoyed the best classification performance in the balancing data when these composite indicators (AUC, F1-Score and G-mean) were used as criteria for model comparison. The values of F1-Score and G-mean for the top three ML models were 0.290/0.660 for Logistic Regression with class weighting, 0.288/0.649 for Stacking with synthetic minority oversampling technique (SMOTE), and 0.285/0.648 for LightGBM with SMOTE. ConclusionsThis paper combining feature selection methods, unbalanced data processing methods and machine learning methods with data from disease surveillance questionnaires and physical measurements to identify people at risk of COPD, concluded that machine learning models based on survey questionnaires could provide an automated identification for patients at risk of COPD, and provide a simple and scientific aid for early identification of COPD.

Power-Expected-Posterior Priors as Mixtures of g-Priors in Normal Linear Models

One of the main approaches used to construct prior distributions for objective Bayes methods is the concept of random imaginary observations. Under this setup, the expected-posterior prior (EPP) offers several advantages, among which it has a nice and simple interpretation and provides an effective way to establish compatibility of priors among models. In this paper, we study the power-expected-posterior prior as a generalization to the EPP in objective Bayesian model selection under normal linear models. We prove that it can be represented as a mixture of g-prior, like a wide range of prior distributions under normal linear models, and thus posterior distributions and Bayes factors are derived in closed form, keeping therefore its computational tractability. Following this result, we can naturally prove that desiderata (criteria for objective Bayesian model comparison) hold for the PEP prior. Comparisons with other mixtures of g-prior are made and results are presented in simulated and real-life datasets.

Modelling inflation dynamics: a Bayesian comparison between GARCH and stochastic volatility

This study employs a prominent model comparison criterion, namely the Bayes factor, to compare three commonly used GARCH models with their stochastic volatility (SV) counterparts in modelling the dynamics of inflation rates. By using consumer price index (CPI) data from 18 developed countries to evaluate these models, we find that the GARCH models are generally outperformed by their stochastic volatility counterparts. Furthermore, the stochastic volatility in mean (SV-M) model is shown to be the best for all 18 countries considered. The paper also examines which model characteristics play a main role in modelling inflation rates. It turns out that inflation volatility feedback is a crucial feature that we should take into consideration when modelling inflation rates. The relevance of a leverage effect, however, is found to be rather ambiguous. Finally, the forecasting results using the log predictive score confirm these findings.

Bayesian inference and prediction in an queueing system

In many real-life situations, certain infinite queueing systems arise in which the clients arrive according to a Poisson process and are served by a single server under an approximately constant service time. An example of such system includes the cycle of a washing machine that takes a fixed length of time to finish one service. This example represents the kind of queueing system that this article attempts to describe, by estimating its traffic intensity ρ, which represents the fraction of time that is busy and can be used to determine other quality measures such as the average queue Lq and the expected number of customers in the system Ls . Bayes estimators for ρ are obtained under the squared error loss function assuming three forms of prior information about ρ, i.e., incomplete gamma prior, left-truncated beta prior, and the improper Jeffreys prior. Furthermore, the Bayes factor as a model comparison criterion is proposed to select a suitable model for Bayesian analysis. A comprehensive set of Monte Carlo simulations comprises our experimental set designed to attest to the efficacy and efficiency of the proposed algorithms. A real case situation is analyzed to illustrate the applicability of the methods developed.

Comparison of M, MM and LTS estimators in linear regression in the presence of outlier

In this study, it was aimed to evaluate the performance of different estimators that will be used in regression analysis, which is one of the multivariate statistical methods in the presence of outliers in the data set. Sixth month live weight was estimated with various body measurements for Saanen kids taken from a private farm. In the data set, the use and performance of robust estimators were evaluated because the least squares method did not provide reliable results in the case of outliers. M (for Huber and Tukey bisquare) estimator, MM estimator and LTS estimator were used as robust used in the presence of outliers. MSE, RMSE, rRMSE, MAPE, MAD, R$^{2}$, R$^{2}$ $_{adj}$ and AIC were used as model comparison criteria in the study. As a result of the study, in the case of outlier in the data set, Huber type M estimator can be recommended.

Comparison of survival models and assessment of risk factors for survival of cardiovascular patients at Addis Ababa Cardiac Center, Ethiopia: a retrospective study.

Cardiovascular diseases (CVDs) is disorders of heart and blood vessels. It is a major health problem across the world, and 82% of CVD deaths is contributed by countries with low and middle income. The aim of this study was to choose appropriate model for the survival of cardiovascular patients data and identify the factors that affect the survival of cardiovascular patients at Addis Ababa Cardiac Center. A Retrospective study was conducted on patients under follow-up at Addis Ababa Cardiac Center between September 2010 to December 2018. The patients included have made either post operation or pre-operation. Out of 1042 cardiac patients, a sample of 332 were selected for the current study using simple random sampling technique. Non-parametric, semi-parametric and parametric survival models were used and comparisons were made to select the appropriate predicting model. Among the sample of 332 cardiac patients, only 67(20.2%) experienced CVD and the remaining 265(79.8%) were censored. The median and the maximum survival time of cardiac patients was 1925 and 1403 days respectively. The estimated hazard ratio of male patients to female patients is 1.926214 (95%CI: 1.111917-3.336847; p = 0.019) implying that the risk of death of male patients is 1.926214 times higher than female cardiac patients keeping the other covariates constant in the model. Even if, all semi parametric and parametric survival models fitted to the current data well, various model comparison criteria showed that parametric/weibull AFT survival model is better than the other. The governmental and non-governmental stakeholders should pay attention to give training on the risk factors identified on the current study to optimize individual's knowledge and awareness so that death due to CVDs can be minimized.

Joint Quantile Regression for Spatial Data

AbstractLinear quantile regression is a powerful tool to investigate how predictors may affect a response heterogeneously across different quantile levels. Unfortunately, existing approaches find it extremely difficult to adjust for any dependency between observation units, largely because such methods are not based upon a fully generative model of the data. For analysing spatially indexed data, we address this difficulty by generalizing the joint quantile regression model of Yang and Tokdar (Journal of the American Statistical Association, 2017, 112(519), 1107–1120) and characterizing spatial dependence via a Gaussian or t-copula process on the underlying quantile levels of the observation units. A Bayesian semiparametric approach is introduced to perform inference of model parameters and carry out spatial quantile smoothing. An effective model comparison criteria is provided, particularly for selecting between different model specifications of tail heaviness and tail dependence. Extensive simulation studies and two real applications to particulate matter concentration and wildfire risk are presented to illustrate substantial gains in inference quality, prediction accuracy and uncertainty quantification over existing alternatives.

Identifying Mixture Components From Large-Scale Keystroke Log Data.

In a computer-based writing assessment, massive keystroke log data can provide real-time information on students’ writing behaviors during text production. This research aims to quantify the writing process from a cognitive standpoint. The hope is that the quantification may contribute to establish a writing profile for each student to represent a student’s learning status. Such profiles may contain richer information to influence the ongoing and future writing instruction. Educational Testing Service (ETS) administered the assessment and collected a large sample of student essays. The sample used in this study contains nearly 1,000 essays collected across 24 schools in 18 U.S. states. Using a mixture of lognormal models, the main findings show that the estimated parameters on pause data are meaningful and interpretable with low-to-high cognitive processes. These findings are also consistent across two writing genres. Moreover, the mixture model captures aspects of the writing process not examined otherwise: (1) for some students, the model comparison criterion favored the three-component model, whereas for other students, the criterion favored the four-component model; and (2) students with low human scores have a wide range of values on the mixing proportion parameter, whereas students with higher scores do not possess this pattern.

Grid search approach to discriminate between old and recent inbreeding using phenotypic, pedigree and genomic information

BackgroundAlthough inbreeding caused by the mating of animals related through a recent common ancestor is expected to have more harmful effects on phenotypes than ancient inbreeding (old inbreeding), estimating these effects requires a clear definition of recent (new) and ancient (old) inbreeding. Several methods have been proposed to classify inbreeding using pedigree and genomic data. Unfortunately, these methods are largely based on heuristic criteria such as the number of generations from a common ancestor or length of runs of homozygosity (ROH) segments. To mitigate these deficiencies, this study aimed to develop a method to classify pedigree and genomic inbreeding into recent and ancient classes based on a grid search algorithm driven by the assumption that new inbreeding tends to have a more pronounced detrimental effect on traits. The proposed method was tested using a cattle population characterized by a deep pedigree.ResultsEffects of recent and ancient inbreeding were assessed on four growth traits (birth, weaning and yearling weights and average daily gain). Thresholds to classify inbreeding into recent and ancient classes were trait-specific and varied across traits and sources of information. Using pedigree information, inbreeding generated in the last 10 to 11 generations was considered as recent. When genomic information (ROH) was used, thresholds ranged between four to seven generations, indicating, in part, the ability of ROH segments to characterize the harmful effects of inbreeding in shorter periods of time. Nevertheless, using the proposed classification method, the discrimination between new and old inbreeding was less robust when ROH segments were used compared to pedigree. Using several model comparison criteria, the proposed approach was generally better than existing methods. Recent inbreeding appeared to be more harmful across the growth traits analyzed. However, both new and old inbreeding were found to be associated with decreased yearling weight and average daily gain.ConclusionsThe proposed method provided a more objective quantitative approach for the classification of inbreeding. The proposed method detected a clear divergence in the effects of old and recent inbreeding using pedigree data and it was superior to existing methods for all analyzed traits. Using ROH data, the discrimination between old and recent inbreeding was less clear and the proposed method was superior to existing approaches for two out of the four analyzed traits. Deleterious effects of recent inbreeding were detected sooner (fewer generations) using genomic information than pedigree. Difference in the results using genomic and pedigree information could be due to the dissimilarity in the number of generations to a common ancestor. Additionally, the uncertainty associated with the identification of ROH segments and associated inbreeding could have an effect on the results. Potential biases in the estimation of inbreeding effects may occur when new and old inbreeding are discriminated based on arbitrary thresholds. To minimize the impact of inbreeding, mating designs should take the different inbreeding origins into consideration.

Chinaʼs Economic Growth, Energy Efficiency, and Industrial Development: Nonlinear Effects on Carbon Dioxide Emissions

This paper analyzes the time-varying impacts of Chinaʼs economic growth, energy efficiency, and industrial development on carbon dioxide (CO2) emissions from 1970 to 2019. First, we examined and found that there are two significant structural changes in the CO2 sequence over the years, and there was a significant nonlinear relationship among the four. The first nonlinear structural model constructed is the TVP regression model. According to the Bayesian model comparison criterion, TVP-SV-VAR was selected as the second constructed model from four types of VAR models containing nonlinear structures. The results show that the conduction intensity value of energy use efficiency to CO2 emissions has increased year by year, from 0.45 in 1971 to 0.97 in 2019. The short-term transmission mechanism of energy use efficiency to carbon emissions is the most significant. The conduction intensity of Chinaʼs economic growth on CO2 emissions increases year by year. Chinaʼs economic growth plays a major role in long-term CO2 emission reduction. The impact of industrial development on CO2 emissions reached a peak of 0.34 in 1977, and the intensity of the impact has basically stabilized at 0.26.

Exploring Class Enumeration in Bayesian Growth Mixture Modeling Based on Conditional Medians

Growth mixture modeling is a popular analytic tool for longitudinal data analysis. It detects latent groups based on the shapes of growth trajectories. Traditional growth mixture modeling assumes that outcome variables are normally distributed within each class. When data violate this normality assumption, however, it is well documented that the traditional growth mixture modeling mislead researchers in determining the number of latent classes as well as in estimating parameters. To address nonnormal data in growth mixture modeling, robust methods based on various nonnormal distributions have been developed. As a new robust approach, growth mixture modeling based on conditional medians has been proposed. In this article, we present the results of two simulation studies that evaluate the performance of the median-based growth mixture modeling in identifying the correct number of latent classes when data follow the normality assumption or have outliers. We also compared the performance of the median-based growth mixture modeling to the performance of traditional growth mixture modeling as well as robust growth mixture modeling based on t distributions. For identifying the number of latent classes in growth mixture modeling, the following three Bayesian model comparison criteria were considered: deviance information criterion, Watanabe-Akaike information criterion, and leave-one-out cross validation. For the median-based growth mixture modeling and t-based growth mixture modeling, our results showed that they maintained quite high model selection accuracy across all conditions in this study (ranged from 87 to 100%). In the traditional growth mixture modeling, however, the model selection accuracy was greatly influenced by the proportion of outliers. When sample size was 500 and the proportion of outliers was 0.05, the correct model was preferred in about 90% of the replications, but the percentage dropped to about 40% as the proportion of outliers increased to 0.15.

Tobacco smoke exposure among women in Turkey and determinants

ABSTRACT Objectives The primary aim of this study was to ascertain the relationship between the demographic and socioeconomic conditions of women aged 15 and older, and their exposure to secondhand smoke (SHS) in Turkey. Method To achieve this, we used the micro dataset of the Turkey Health Survey conducted by the Turkish Statistics Institute (TurkStat), specifically the data from 19,982 women aged 15 or older. We performed our analysis with ordered logistic regression, ordered probit regression, generalized ordered logistic regression, and generalized ordered probit regression to determine the factors related to individual exposure to tobacco smoke. Results According to the model comparison criteria, generalized ordered logistic regression produced the best model. The consequent results of this model indicated that age, education level, working conditions, average monthly net income, general health status, psychosocial support conditions, alcohol consumption, and the respective survey year were efficient variables with regard to tobacco smoke exposure. Conclusion This study highlights, in particular, the need to target young employed women of low socioeconomic and educational levels, who consume alcohol, with poor overall health, and receive psychosocial support. These groups are more vulnerable to SHS exposure, thus tobacco prevention and quitting programs should focus on these groups.

A Bayesian marked spatial point processes model for basketball shot chart

Abstract The success rate of a basketball shot may be higher at locations where a player makes more shots. For a marked spatial point process, this means that the mark and the intensity are associated. We propose a Bayesian joint model for the mark and the intensity of marked point processes, where the intensity is incorporated in the mark model as a covariate. Inferences are done with a Markov chain Monte Carlo algorithm. Two Bayesian model comparison criteria, the Deviance Information Criterion and the Logarithm of the Pseudo-Marginal Likelihood, were used to assess the model. The performances of the proposed methods were examined in extensive simulation studies. The proposed methods were applied to the shot charts of four players (Curry, Harden, Durant, and James) in the 2017–2018 regular season of the National Basketball Association to analyze their shot intensity in the field and the field goal percentage in detail. Application to the top 50 most frequent shooters in the season suggests that the field goal percentage and the shot intensity are positively associated for a majority of the players. The fitted parameters were used as inputs in a secondary analysis to cluster the players into different groups.