Search

This study utilized Chlorophyll-a, sea surface temperature (SST), and sea surface height (SSH) as the environmental variables to identify skipjack tuna catch hotspots. This study conducted statistical methods (decision tree, DT, and generalized linear model, GLM) as ensemble models that were employed for predicting skipjack area for each time slice. Using spatial historical data, each model was trained for one of the ensemble model sets. For prediction, the correlations of historical and new inputs were applied to select the predictive model. Using the scene-based model with the highest input correlation, this study further identified the fishing area of skipjack tuna in every case whether the alterations in their environment affected their abundance or not. Overall, the performance achieved over 83% for correlation coefficients (CC) based on the accuracy assessment. This study concluded that DT appears to perform better than GLM in predicting skipjack tuna fishing areas. Moreover, the most influential environmental variable in model construction was sea surface temperature (SST), indicating that the presence of skipjack tuna was primarily influenced by regional temperature.

To study the relationship between genetic variants and phenotypes, association testing is adopted; however, most association studies are conducted by genotype-based testing. Testing methods based on next-generation sequencing (NGS) data without genotype calling demonstrate an advantage over testing methods based on genotypes in the scenarios when genotype estimation is not accurate. Our objective was to develop NGS data-based methods for association studies to fill the gap in the literature. Single-variant testing methods based on NGS data have been proposed, including our previously proposed single-variant NGS data-based testing method, i.e., UNC combo method. The NGS data-based group testing method has been proposed by us using a linear model framework which can handle continuous responses. In this paper, we extend our linear model-based framework to a generalized linear model-based framework so that the methods can handle other types of responses especially binary responses which is a common problem in association studies. To evaluate the performance of various estimators and compare them we performed simulation studies. We found that all methods have Type I errors controlled, and our NGS data-based methods have better performance than genotype-based methods for other types of responses, including binary responses (logistics regression) and count responses (Poisson regression), especially when sequencing depth is low. We have extended our previous linear model (LM) framework to a generalized linear model (GLM) framework and derived NGS data-based methods for a group of genetic variables. Compared with our previously proposed LM-based methods, the new GLM-based methods can handle more complex responses (for example, binary responses and count responses) in addition to continuous responses. Our methods have filled the literature gap and shown advantage over their corresponding genotype-based methods in the literature.

This paper focuses on estimation and null region detection of the coefficient function for functional generalized linear models. Traditional estimating approaches can not serve as null regions detectors. To simultaneously estimate coefficient functions and detect corresponding important subregions in functional generalized linear models, an adaptive group LASSO approach with B-spline smoothing technique is developed. The convergence rate of the resulting estimator is obtained. The consistency of domain selection and limiting distribution of the proposed estimator are then established, which is not straightforward since the groups to be penalized are overlapping. These asymptotic properties are also supported by extensive simulation studies. The resulting estimator performs better than direct adaptive LASSO estimators and some existing functional generalized linear models estimators, which are obtained without considering the domain selection. A real data application reveals the effectiveness of the proposed method.

Background: Desired family size, the number of children wanted in one’s lifetime, is one of the major factors influencing the population dynamics. Knowledge of factors influencing the desired family size is crucial in regulating the population growth. This study assesses the desired family size and its determinants among Ethiopian women. Methods: The study was based on the 2016 Ethiopian Demographic and Health Survey (EDHS) data. A total of 13,941 women with complete information were considered in the study. Count regression models, a family of generalized linear models, were used to identify determinant factors for the desired number of children among women in Ethiopia. Results: Results showed that the average number of children per women was 4.5. The highest average was recorded in Somali region (10.58) and the lowest in Addis Ababa city administration (3.56). The random intercept negative binomial regression model was found to be the most appropriate model for the data. Women’s age, household head, contraception using, wealth index, women’s education, religion, marital status, husband’s work, family size and age at first marriage were significantly associated with the desired number of children by women at reproductive age. The study also showed that there was significant regional variation in desired number of children among Ethiopian women. Conclusions: Variables such as women’s age and family size were positively associated with the desired number of children by women at reproductive age while age at first marriage, contraception using, wealth index and women’s education were negatively associated with the desired number of children by women. Women who attained higher level of education preferred small family size compared to women with no formal education. Desire for children was high among mothers from rural areas, women in the poorest economic level and woman having no work. Keywords: Desired number of children, statistical weight, EDHS, Count regression models, multilevel analysis.

Background Inference using standard linear regression models (LMs) relies on assumptions that are rarely satisfied in practice. Substantial departures, if not addressed, have serious impacts on any inference and conclusions; potentially rendering them invalid and misleading. Count, bounded and skewed outcomes, common in physical activity research, can substantially violate LM assumptions. A common approach to handle these is to transform the outcome and apply a LM. However, a transformation may not suffice. Methods In this paper, we introduce the generalized linear model (GLM), a generalization of the LM, as an approach for the appropriate modelling of count and non-normally distributed (i.e., bounded and skewed) outcomes. Using data from a study of physical activity among older adults, we demonstrate appropriate methods to analyse count, bounded and skewed outcomes. Results We show how fitting an LM when inappropriate, especially for the type of outcomes commonly encountered in physical activity research, substantially impacts the analysis, inference, and conclusions compared to a GLM. Conclusions GLMs which more appropriately model non-normally distributed response variables should be considered as more suitable approaches for managing count, bounded and skewed outcomes rather than simply relying on transformations. We recommend that physical activity researchers add the GLM to their statistical toolboxes and become aware of situations when GLMs are a better method than traditional approaches for modeling count, bounded and skewed outcomes.

Objective: Generalized linear models (GLMs) such as logistic and Poisson regression are among the most common statistical methods for modeling binary and count outcomes. Though single-coefficient tests (odds ratios, incidence rate ratios) are the most common way to test predictor-outcome relations in these models, they provide limited information on the magnitude and nature of relations with outcomes. We assert that this is largely because they do not describe direct relations with quantities of interest (QoIs) such as probabilities and counts. Shifting focus to QoIs makes several nuances of GLMs more apparent.Method: To bolster interpretability of these models, we provide a tutorial on logistic and Poisson regression and suggestions for enhancements to current reporting practices for predictor-outcome relations in GLMs.Results: We first highlight differences in interpretation between traditional linear models and GLMs, and describe common misconceptions of GLMs. In particular, we highlight that link functions A) introduce non-constant relations between predictors and outcomes and B) make predictor-QoI relations dependent on other covariates. Each of these properties causes interpretation of GLM coefficients to diverge from interpretations of linear models. Next, we argue for a more central focus on QoIs (probabilities and counts). Finally, we propose and provide graphics and tables, with sample R code, for enhancing presentation and interpretation of QoIs.Conclusions: By improving present practices in the reporting of predictor-outcomerelations in GLMs, we hope to maximize the amount of actionable information generated by statistical analyses and provide a tool for building a cumulative science of substance use disorders.Public Health Significance: We propose several enhancements to current reporting practices for statistical analyses of binary outcomes (e.g., psychiatric diagnoses) and count outcomes (e.g., number of alcoholic drinks consumed). We encourage researchers to interpret results in terms of quantities of interest (probabilities for binary models, counts for count models) and provide a tutorial and R code for implementing these analyses. Doing so can provide richer information about a statistical analysis, make study results easier for research consumers to understand, and facilitate comparison of results across studies.

In Indonesia, rainfall is one crucial triggering factor for landslides. This paper aims to build landslide event prediction models using several machine learning and artificial intelligence algorithms. The algorithms were trained with two different methods. The input of the algorithms was precipitation data obtained from the global satellite mapping of precipitation satellite observation, and the target was landslide event occurrence data obtained from the Indonesian National Board for Disaster Management. Each algorithm provided some model candidates with different parameter settings for each method. As a result, there were 52 and 72 model candidates for both methods. The best model was then chosen from each method. The result shows that the model generated by generalized linear model was the best model for the first method and deep learning for the second one. Furthermore, the best models at each method gained 0.828 and 0.836 for the area under receiver operating characteristics curve, and their log-loss were 0.156 and 0.154. The second method, which used input data transformation, provided better performance.

Adding domain knowledge to a learning system is known to improve results. In multiparameter Bayesian frameworks, such knowledge is incorporated as a prior. On the other hand, the various model parameters can have different learning rates in real-world problems, especially with skewed data. Two often-faced challenges in operation management and management science applications are the absence of informative priors and the inability to control parameter learning rates. In this study, we propose a hierarchical empirical Bayes approach that addresses both challenges and that can generalize to any Bayesian framework. Our method learns empirical meta-priors from the data itself and uses them to decouple the learning rates of first-order and second-order features (or any other given feature grouping) in a generalized linear model. Because the first-order features are likely to have a more pronounced effect on the outcome, focusing on learning first-order weights first is likely to improve performance and convergence time. Our empirical Bayes method clamps features in each group together and uses the deployed model’s observed data to empirically compute a hierarchical prior in hindsight. We report theoretical results for the unbiasedness, strong consistency, and optimal frequentist cumulative regret properties of our meta-prior variance estimator. We apply our method to a standard supervised learning optimization problem as well as an online combinatorial optimization problem in a contextual bandit setting implemented in an Amazon production system. During both simulations and live experiments, our method shows marked improvements, especially in cases of small traffic. Our findings are promising because optimizing over sparse data is often a challenge. This paper was accepted by Hamid Nazerzadeh, Management Science Special Section on Data-Driven Prescriptive Analytics.

Climate finance and carbon pricing are regarded as sustainable policy mechanisms for mitigating negative environmental externalities via the development of green financing projects and the imposition of taxes on carbon pollution generation. Financial literacy indicates that it is beneficial to invest in cleaner technology to advance the environmental sustainability goal. The current wave of the COVID-19 epidemic has had a detrimental effect on the world economies' health and income. The pandemic crisis dwarfs previous global financial crises in terms of scope and severity, collapsing global financial markets. The study's primary contribution is constructing a climate funding index (CFI) based on four critical factors: inbound foreign direct investment, renewable energy usage, research and development spending, and carbon damages. In a cross-sectional panel of 43 nations, the research evaluates the effect of climate funding, financial literacy, and carbon pricing in lowering exposure to coronavirus cases. The study utilized Newton-Raphson and Marquardt steps to estimate the current parameter estimates while evaluating the COVID-19 prediction model with level regressors using the robust least squares regression model (S-estimator). Additionally, the innovation accounting matrix predicts estimations over a specific period. The findings indicate that climate finance significantly reduces coronavirus exposure by introducing green financing initiatives that benefit human health, which eventually strengthens the immune system's ability to fight infectious illnesses. Financial literacy and carbon pricing, on the other hand, are ineffectual in controlling coronavirus infections due to rising economic activity and densely inhabited areas that enable the transmission of coronavirus cases across countries. Similar findings were obtained using the alternative regression apparatus. The COVID-19 predicted variable was used as a "response variable," and climate financing was shown to have a favorable impact on containing coronavirus exposure. As shown by the innovation accounting matrix, carbon pricing would drastically decrease coronavirus cases' exposure over a time horizon. The study concludes that climate finance and carbon pricing were critical in improving air quality indicators, which improved countries' health and wealth, allowing them to reduce coronavirus infections via sustainable healthcare reforms.

Floods are among the devastating natural disasters that occurred very frequently in arid regions during the last decades. Accurate assessment of the flood susceptibility mapping is crucial in sustainable development. It helps respective authorities to prevent as much as possible their irreversible consequences. The Digital Elevation Model (DEM) spatial resolution is one of the most crucial base layer factors for modeling Flood Probability Maps (FPMs). Therefore, the main objective of this study was to assess the influence of the spatial resolution of the DEMs 12.5 m (ALOS PALSAR) and 30 m (ASTER) on the accuracy of flood probability prediction using three machine learning models (MLMs), including Random Forest (RF), Artificial Neural Network (ANN), and Generalized Linear Model (GLM). This study selected 14 causative factors in the flood as independent variables, and 220 flood locations were selected as dependent variables. Dependent variables were divided into training (70%) and validation (30%) for flood susceptibility modeling. The Receiver Operating Characteristic Curve (ROC), Kappa index, accuracy, and other statistical criteria were used to evaluate the models' accuracy. The results showed that resolving the DEM alone cannot significantly affect the accuracy of flood probability prediction regardless of the applied MLM and independently of the statistical model used to assess the performance accuracy. In contrast, the factors such as altitude, precipitation, and distance from the river have a considerable impact on floods in this region. Also, the evaluation results of the models showed that the RF (AUC12.5,30m = 0.983, 0.975) model is more accurate in preparing the FPM than the ANN (AUC12.5,30m = 0.949, 0.93) and GLM (AUC12.5,30m = 0.965, 0.949) models. This study's solution-oriented findings might help water managers and decision-makers to make the most effective adaptation and mitigation measures against potential flooding.

Nowadays, life is intimately associated with transportation, generating several issues on it. Numerous works are available concerning accident prediction techniques depending on independent road and traffic features, while the mix parameters including time, geometry, traffic flow, and weather conditions are still rarely ever taken into consideration. This study aims to predict future accident frequency and the risk factors of traffic accidents. It utilizes the Generalized Linear Model (GLM) and Artificial Neural Networks (ANN) approaches to process and predict traffic data efficiently based on 21500 records of traffic accidents that occurred in Erzurum in Turkey from 2005 to 2019. The results of the comparative evaluation demonstrated that the ANN model outperformed the GLM model. The study revealed that the most effective variable was the number of horizontal curves. The annual average growth rates of accident occurrences based on the ANNꞌs method are predicted to be 11.22% until 2030.

Background Heterogeneity in single-cell RNA-seq (scRNA-seq) data is driven by multiple sources, including biological variation in cellular state as well as technical variation introduced during experimental processing. Deconvolving these effects is a key challenge for preprocessing workflows. Recent work has demonstrated the importance and utility of count models for scRNA-seq analysis, but there is a lack of consensus on which statistical distributions and parameter settings are appropriate. Results Here, we analyze 59 scRNA-seq datasets that span a wide range of technologies, systems, and sequencing depths in order to evaluate the performance of different error models. We find that while a Poisson error model appears appropriate for sparse datasets, we observe clear evidence of overdispersion for genes with sufficient sequencing depth in all biological systems, necessitating the use of a negative binomial model. Moreover, we find that the degree of overdispersion varies widely across datasets, systems, and gene abundances, and argues for a data-driven approach for parameter estimation. Conclusions Based on these analyses, we provide a set of recommendations for modeling variation in scRNA-seq data, particularly when using generalized linear models or likelihood-based approaches for preprocessing and downstream analysis.

The aim of this study was to evaluate the effect of gestational weight gain (GWG) up to 23+6 weeks of gestation on the incidence of Gestational Diabetes Mellitus (GDM). A pregnant cohort of 5948 women in Northern Greece was recruited. Anthropometric features before and during pregnancy were recorded, the GWG by 23+6 weeks was calculated and a Generalized Linear Regression Model (GLM) with subgroup analyses based on weight status were computed. GDM was diagnosed in 5.5% of women. GLM results showed that GDM likelihood increased with maternal age (MA) and pre-pregnancy BMI (aOR: 1.08, 95%CI: [1.06, 1.11] and aOR: 1.09, 95%CI: [1.09, 1.11], respectively). Ιn the normal pre-pregnancy weight group, when the extra weight gain was >8 kgs, the odds of GDM increased (OR: 2.13, 95%CI: [0.98, 4.21], p = 0.03). Women with pre-pregnancy level 2 clinical obesity (OB2 pre) (BMI > 35 and

The following paper presents as a research problem the ethnic-regional differences in the allocation of high complexity spending in Brazil in an analysis from 2010 to 2019. This is a descriptive research in which a generalized linear model (GLM) was developed to analyze these hospital expenditures with high complexity procedures. The total spending on high complexity procedures in Brazil has increased over the past decade. The study shows that the lowest average expenditures are found in the North and Northeast regions. When comparing the spending between different ethnicities, it was observed that the only decrease between the years 2010 and 2019 was in the amount spent on procedures in indigenous people. The spending on male patients was significantly higher compared to female patients. The highest expenditures, on the other hand, are concentrated in the regions of state capitals favoring the strengthening of hub municipalities. Geographic inequalities in access still persist, even with most states already offering almost all procedures. The Brazilian territory is very heterogeneous and needs to organize its health system by regions, therefore integrated public policies and economic and social development are urgently needed.

El documento es una investigación para optar por el grado de Magíster en Ingeniería Civil, mención Transportes, por parte de Alvaro Amilcar Soto Espinoza (Cod. 201610452). En esta se estudia y comenta la relación entre las estaciones de transporte público masivo y los siniestros viales ocurridos en el territorio entorno a ellas, para el caso de la conurbación de Lima y Callao, Perú. El estudio de la relación entre las estaciones de transporte público masivo y la siniestralidad vial ocasionada alrededor de ellas ha sido somero en la literatura internacional y nulo en la local. En respuesta, la investigación explora esta relación para la conurbación de Lima y Callao, Perú, en base a datos correspondientes al quinquenio 2015-2019, para el conjunto de 38 estaciones de la línea de buses de tránsito rápido y 26 de la línea de metro. Como resultado, se ha determinado una asociación significativa y directa entre la ocurrencia de los eventos fatales y la presencia conjunta de ciertos grandes centros atractores de viajes, poblaciones etariamente vulnerables, y barrios de estrato bajo, bajo medio y medio. A partir de estos hallazgos y otros secundarios se propone un Modelo Lineal Generalizado de Poisson Multivariable para la estimación de los siniestros viales fatales, así como múltiples aprendizajes y recomendaciones para las entidades públicas a cargo de la toma de decisiones. The study of the relationship between mass public transport stations and road crashes on its surroundings has been little by the international community and null by local. In response, the investigation explores this issue for the conurbation of Lima and Callao, Peru, based on data for the period 2015-2019, for the sum of 38 one-line bus rapid transit stations and 26 another-line metro stations. As result, it was found a significant direct relation between fatal road crashes and presence of some main-destination centres, young and elderly population, and low- and middle-income neighbourhoods. Based on these findings and other ones a Poisson Multivariable Generalized Linear Model for fatal road crash estimation was proposed, as well as learnings and recommendations for public decision makers. Magíster en Ingeniería Civil Maestría

In applications of generalized linear models to education, observations at one level are frequently nested within units at another level. For example, we have measurements on students, who are located within a class, which is within a school, and so on. Methods for analyzing such data go back several decades, but new methods have led to many extensions and interesting ways of interpreting such models. These include models for multiple measures on individuals (such as longitudinal studies), and individuals nested within studies (for meta-analysis). These models also directly solve the aggregation problem: statistics on higher level units (e.g., correlations of school or classroom means) often do not equal (and may even be opposite in sign to) those for lower level units (both overall and within each higher level unit).

Terrestrial gastropods are among the most imperiled groups of organisms on Earth. Many species have a complex taxonomic history, often including poorly defined subspecies, most of which have not been the focus of modern systematics research. Genomic tools, geometric morphometrics, and environmental niche modeling were used to assess the taxonomic status of Patera clarkii nantahala (Clench & Banks, 1932), a subspecies of high conservation concern with a restricted range of approximately 3.3 km2 in North Carolina, USA. A genome-scale dataset was generated that included individuals with morphologies matching P. c. nantahala, P. c. clarkii, and one individual with an intermediate form between P. c. nantahala and P. c. clarkii that was initially hypothesized as a potential hybrid. Mitochondrial phylogenetics, nuclear species tree inference, and phylogenetic networks were used to assess relationships and gene flow. Differences in shell shape via geometric morphometrics and whether the environmental niches of the two subspecies were significantly different were also examined. Molecular analyses indicated an absence of gene flow among lineages of P. clarkii sensu lato. Analyses rejected our hypothesis that the intermediate shelled form represented a hybrid, but instead indicated that it was a distinct lineage. Environmental niche models indicated significant differences in environmental niche between P. c. clarkii and P. c. nantahala, and geometric morphometrics indicated that P. c. nantahala had a significantly different shell shape. Given multiple lines of evidence, species-level recognition of P. nantahala is warranted.

Jinlian Jin, Haiyan Zhou, Shulin Sun, Zhe Tian, Haibing Ren, Jinwu Feng Department of Gastroenterology, The Third Clinical Medical College of China Three Gorges University, Gezhouba Central Hospital of Sinopharm, Yichang, Hubei, 443002, People’s Republic of ChinaCorrespondence: Jinlian JinDepartment of Gastroenterology, The Third Clinical Medical College of China Three Gorges University, Gezhouba Central Hospital of Sinopharm, No. 60, Qiaohu 1st Road, Xiling District, Yichang, Hubei, 443002, People’s Republic of ChinaTel +8613986746553Email jjl7475@163.comPurpose: Predicting lymph node metastasis (LNM) after endoscopic resection is crucial in determining whether patients with pT1NxM0 colorectal cancer (CRC) should undergo additional surgery. This study was aimed to develop a predictive model that can be used to reduce the current likelihood of overtreatment.Patients and Methods: We recruited a total of 1194 consecutive CRC patients with pT1NxM0 who underwent endoscopic or surgical resection at the Gezhouba Central Hospital of Sinopharm between January 1, 2006, and August 31, 2021. The random forest classifier (RFC) and generalized linear algorithm (GLM) were used to screen out the variables that greatly affected the LNM prediction, respectively. The area under the curve (AUC) and decision curve analysis (DCA) were applied to assess the accuracy of predictive models.Results: Analysis identified the top 10 candidate factors including depth of submucosal invasion, neutrophil-lymphocyte ratio (NLR), platelet lymphocyte ratio (PLR), platelet-to-neutrophil ratio(PNR), venous invasion, poorly differentiated clusters, tumor budding, grade, lymphatic vascular invasion, and background adenoma. The performance of the GLM achieved the highest AUC of 0.79 (95% confidence interval [CI]: 0.30 to 1.28) in the training cohort and robust AUC of 0.80 (95% confidence interval [CI]: 0.36 to 1.24) in the validation cohort. Meanwhile, the RFC exhibited a robust AUC of 0.84 (95% confidence interval [CI]: 0.40 to 1.28) in the training cohort and a high AUC of 0.85 (95% CI: 0.41 to 1.29) in the validation cohort. DCAs also showed that the RFC had superior predictive ability.Conclusion: Our supervised learning-based model incorporating histopathologic parameters and inflammatory markers showed a more accurate predictive performance compared to the GLM. This newly supervised learning-based predictive model can be used to determine an individually tailored treatment strategy.Keywords: colorectal cancer, pT1NxM0, lymph nodes metastasis, prediction model, machine learning, random forest classifier, generalized linear model

Bike Sharing Systems (BSS) have gained popularity all over the world. BSS success requires efficient location of stations and the rebalancing of the network, both of which depend on predicted demand and flows between stations. Therefore, predicting BSS demand is essential to optimize logistic operations, as well as network reliability and availability. The present work develops statistical models to predict the hourly origin-destination demand in the Lisbon BSS network. Models predict the number of trips that will occur between an origin station and a destination station (OD pair) at a certain hour, as well as the duration of such trips. This study also examines the effect of several explaining variables, including the weather, intermodality effects, effects of station location close to different types of points of interest and the time/hour effect. These effects were all tested using different statistical models, including the Generalized Linear Models (GLM) and zero-augmented ones: the hurdle model and the zero-inflated model. The hurdle model showed the best fit to the data and several explaining variables exhibited statistical significant effects. A discussion on the potential logistic and policy impacts of such findings is also provided, so that BSS network design and operations can incorporate them within the urban mobility system policy.

Two commonly used methods for improving precision and power in clinical trials are stratified randomization and covariate adjustment. However, many trials do not fully capitalize on the combined pr...

A natural dependence among diameters measured within-tree is expected in taper data due to the hierarchical structure. The aim of this paper was to introduce the covariance generalized linear model (CGLM) framework in the context of forest biometrics for Pinus taeda stem form modeling. The CGLMs are based on marginal specification, which requires a definition of the mean and covariance components. The tree stem mean profiles were modeled by a nonlinear segmented model. The covariance matrix was built considering four strategies of linear combinations of known matrices, which expressed the variance or correlations among observations. The first strategy modeled only the variance of the diameters over the stem as a function of covariates, the second modeled correlation among observations, the third was defined based on a random walk model, the fourth was based on a structure similar to a mixed-effect model with a marginal specification, and the fourth was a traditional mixed-effect model. Mean squared error and bias showed that the approaches were similar for describing the mean profile for fitting and validation dataset. However, uncertainties expressed by confidence intervals of the relative diameters were significant and related to the matrix covariance structures of the CGLMs.

Nitrate is a major pollutant in groundwater whose main source is municipal wastewater and agricultural activities. In the present study, Bayesian approaches such as Bayesian generalized linear model (BGLM), Bayesian regularized neural network (BRNN), Bayesian additive regression tree (BART), and Bayesian ridge regression (BRR) were used to model groundwater nitrate contamination in a semiarid region Marvdasht watershed, Fars province, Iran. Eleven groundwater (GW) nitrate conditioning factors have been taken as input parameters for predictive modeling. The results showed that the Bayesian models used in this study were all competent to model groundwater nitrate and the BART model with R2 = 0.83 was more efficient than the other models. The result of variable importance showed that potassium (K) has the highest importance in the models followed by rainfall, altitude, groundwater depth, and distance from the residential area. The results of the study can support the decision-making process to control and reduce the sources of nitrate pollution.

Social media platforms have become accessible resources for health data analysis. However, the advanced computational techniques involved in big data text mining and analysis are challenging for public health data analysts to apply. This study proposes and explores the feasibility of a novel yet straightforward method by regressing the outcome of interest on the aggregated influence scores for association and/or classification analyses based on generalized linear models. The method reduces the document term matrix by transforming text data into a continuous summary score, thereby reducing the data dimension substantially and easing the data sparsity issue of the term matrix. To illustrate the proposed method in detailed steps, we used three Twitter datasets on various topics: autism spectrum disorder, influenza, and violence against women. We found that our results were generally consistent with the critical factors associated with the specific public health topic in the existing literature. The proposed method could also classify tweets into different topic groups appropriately with consistent performance compared with existing text mining methods for automatic classification based on tweet contents.

The manufacturing industry is an important pillar of the national economy. It is of vital importance to develop statistical modellings in order to quantify the relationship between potential internal drivers and the trend of output values in the manufacturing industry. However, only a few statistical modellings have been established to investigate such associations. This study developed the correlation coefficient model and generalized linear model (GLM) to measure the single and interactive effects of the internal drivers on the changes of the output values. For the GLM, different predictive variables were developed to fit into the dataset, and the performance of the models were compared using fitness parameters. Furthermore, an industry survey dataset for 1,180 manufacturing enterprises in 2020 was used to validate the models. The use of the GLM combining land area, number of employees, scientific research input, and labor productivity may have a great potential to bolster capacity in monitoring and predicting the trend of output values in the manufacture industry.

Habitat suitability modeling and mapping are important aspects of long-term strategies for sustaining plant ecosystems. In this study, seven state-of-the-art machine learning models including boosted regression tree (BRT), functional discriminant analysis (FDA), generalized linear model (GLM), multivariate adaptive regression splines (MARS), mixture discriminant analysis (MDA), random forest (RF), and support vector machine (SVM) were applied to model habitat suitability for Ferula gummosa medicinal plant in the Firozkuh County of Tehran. Different factors that affect the habitat of this plant were prepared for modeling, including slope angle, silt percentage, sand percentage, aspect, annual mean rainfall, clay percentage, topographic wetness index, elevation, distance from rivers, drainage density, annual mean temperature, plan curvature, profile curvature, land use, lithological units, and organic carbon. After running the models in R software, their evaluation using various measures (area under the curve, accuracy, precision, F-measure, fallout, true skill statistics, and corrected classify instances) indicated that the RF model was the best one for assessing Ferula gummosa habitat suitability. The SVM, MARS, MDA, GLM, FDA, and BRT models also displayed acceptable performances. The results of our study contribute to the understanding of the stability of the medicinal plant Ferula gummosa and to help avoid its extinction in the future.

We consider the following question: given information on individual policyholder characteristics, how can we ensure that insurance prices do not discriminate with respect to protected characteristics, such as gender? We address the issues of direct and indirect discrimination, the latter resulting from implicit learning of protected characteristics from nonprotected ones. We provide rigorous mathematical definitions for direct and indirect discrimination, and we introduce a simple formula for discrimination-free pricing, that avoids both direct and indirect discrimination. Our formula works in any statistical model. We demonstrate its application on a health insurance example, using a state-of-the-art generalized linear model and a neural network regression model. An important conclusion is that discrimination-free pricing in general requires collection of policyholders’ discriminatory characteristics, posing potential challenges in relation to policyholder’s privacy concerns., ASTIN Bulletin, 52 (1), ISSN:0515-0361, ISSN:1783-1350

A statistical profile is a relationship between a quality characteristic (a response) and one or more explanatory variables to characterize quality of a process or a product. Monitoring profiles or checking the stability of profiles over time, has been extensively studied under the normal response variable, but it has paid a little attention to the profile with the non-normal response variable denoted by generalized linear models (GLM). Whereas, some of the potential applications of profile monitoring are cases where the response can be modelled using logistic profiles entailing binary, nominal and ordinal models. Also, most of existing control charts in this field have been developed by statistical approach and employing machine learning techniques have been rarely addressed in the related literature. Hence, to implement on-line process monitoring of logistic profiles, a novel artificial neural network (ANN) as a control chart with a heuristic training procedure is proposed in this paper. Performance of the proposed approach is investigated and compared using simulation studies in binary and polytomous models based on average run length (ARL) criterion. Simulation results revealed a good performance of the proposed approach. Nevertheless, to enhance the detection ability of the proposed approach more, the idea of combining run-rule which is a supplementary tool for making more sensitive control chart with final statistic is also implemented in this paper. Furthermore, a diagnostic method with machine learning schemes is employed to identify the shifted parameters in the profile. Results indicate the superior performance of the proposed approaches in most of the simulations. Finally, an example is used to illustrate the implementation of the proposed charting scheme.

The identification of influential observations is an essential element in regression analysis as they posed a threat to the model building process. The existence of multicollinearity among the regressors complicates the presence of influential observations. Different influential diagnostics have been presented in literature so far using generalized linear models (GLM). In this paper, approximate deletion measures based on Sherman–Morrison Woodbury (SMW) theorem for the Poisson Two-Parameter regression model are proposed to detect influential observations in the presence of multicollinearity. Moreover, we conduct a Monte Carlo Simulation to evaluate the performance of the proposed measures. Finally, an example is presented to illustrate the proposed diagnostic measures.

It is important to understand how animals respond to changes in landscape structure, especially when considering habitat alteration and urbanization. Using the burrowing owl (Athene cunicularia) as an ecological model we tested two hypotheses: (1) the landscape of urban areas influences trajectories of the burrowing owl, which responds changing its displacement patterns, and (2) models incorporating species-specific traits, such as stochasticity and landscape perception, better predict flight trajectories than least-cost path models. Thirty owls were captured at the Brasilia International Airport, fitted with VHF transmitters, and released in random locations within 30 km of the airport. We generated a habitat suitability map for use as a resistance raster to model flight trajectories: one model was based on the least-cost path, and four models included stochasticity and variations of the owls’ landscape perception and dependency on natural ecosystems. Observed trajectories were compared with those predicted by models using the following metrics extracted from the trajectories: mean habitat suitability values, sinuosity, and length. The best generalized linear models were selected using the Akaike information criterion. The owls dispersed through areas with a lower cost than expected by chance (z = 104.65, p 0.05), sinuosity (r2 = 0.03, p > 0.05), and length (r2 = 0.37, p > 0.05). Our results demonstrate that landscape information influences flight trajectories, and models incorporating landscape perception and stochasticity better predict burrowing owl flight trajectories.

We discuss the modeling of temporal dominance of sensations (TDS) data, time series data appearing in sensory analysis, that describe temporal changes of the dominant taste in the oral cavity. Our aims were to obtain the transition process of attributes (tastes and mouthfeels) in the oral cavity, to express the tendency of dominance durations of attributes, and to specify factors (such as sex, age, food preference, dietary habits, and sensitivity to a particular taste) affecting dominance durations, simultaneously. To achieve these aims, we propose an analysis procedure applying models based on the semi-Markov chain and the negative binomial regression, one of the generalized linear models. By using our method, we can take differences among individual panelists and dominant attributes into account. We analyzed TDS data for milk chocolate with the proposed method and verified the performance of our model compared with conventional analysis methods. We found that our proposed model outperformed conventional ones; moreover, we identified factors that have effects on dominance durations. Results of an experiment support the importance of reflecting characteristics of panelists and attributes.The online version contains supplementary material available at 10.1007/s13197-021-05260-9.

The beta prime regression model proposed by Bourguignon, Santos-Neto, and de Castro is an alternative to the generalized linear models and useful to model positive asymmetric data. In this paper, w...

We propose a nonparametric bivariate varying coefficient generalized linear model to predict a mean response trajectory in the future given an individual's characteristics at present or an earlier time point in a longitudinal study. Given the measurement time of the predictors, the coefficients vary as functions of the future time over which the prediction of the mean response is concerned and illustrate the dynamic association between the future response and the earlier measured predictors. We use a nonparametric approach that takes advantage of features of both the kernel and the spline methods for estimation. The resulting coefficient estimator is asymptotically consistent under mild regularity conditions. We also develop a new bootstrap approach to construct simultaneous confidence bands for statistical inference about the coefficients and the predicted response trajectory based on the coverage rate of bootstrap estimates. We use the Framingham Heart Study to illustrate the methodology. The proposed procedure is applied to predict the probability trajectory of hypertension risk given individuals' health condition in early adulthood and to examine the impact of risk factors in early adulthood on a long-term risk of hypertension over several decades.

The approximation theory is utilized in optimal design finding by approximating the covariance matrix by the Fisher information matrix; however, the approximation accuracy depends on the sample siz...

Mosquitoes transmit several infectious diseases that pose significant threat to human health. Temperature along with other environmental factors at breeding and resting locations play a role in the organismal development and abundance of mosquitoes. Accurate analysis of mosquito population dynamics requires information on microclimatic conditions at breeding and resting locations. In this study, we develop a regression model to characterize microclimatic temperature based on ambient environmental conditions. Data were collected by placing sensor loggers at resting and breeding locations such as storm drains across Houston, TX. Corresponding weather data was obtained from National Oceanic and Atmospheric Administration website. Features extracted from these data sources along with contextual information on location were used to develop a Generalized Linear Model for predicting microclimate temperatures. We also analyzed mosquito population dynamics for Aedes albopictus under ambient and microclimatic conditions using system dynamic (SD) modelling to demonstrate the need for accurate microclimatic temperatures in population models. The microclimate prediction model had an R2 value of ~ 95% and average prediction error of ~ 1.5 °C indicating that microclimate temperatures can be reliably estimated from the ambient environmental conditions. SD model analysis indicates that some microclimates in Texas could result in larger populations of juvenile and adult Aedes albopictus mosquitoes surviving the winter without requiring dormancy.

In this study, investigation of the economic growth of the Organization for Economic Cooperation and Development (OECD) countries and the countries in different income groups in the World Data Bank is conducted by using causality analyses and Generalized Estimating Equations (GEEs) which is an extension of Generalized Linear Models (GLMs). Eight different macro-economic, energy and environmental variables such as the gross domestic product (GDP) (current US$), CO2 emission (metric tons per capita), electric power consumption (kWh per capita), energy use (kg of oil equivalent per capita), imports of goods and services (% of GDP), exports of goods and services (% of GDP), foreign direct investment (FDI) and population growth rate (annual %) have been used. These countries have been categorized according to their OECD memberships and income groups. The causes of the economic growth of these countries belonging to their OECD memberships and income groups have been determined by using the Toda-Yamamoto causality test. Furthermore, various GEE models have been established for the economic growth of these countries belonging to their OECD membership and income groups in the aspect of the above variables. These various GEE models for the investigation of the economic growth of these countries have been compared to examine the contribution of the causality analyses to the statistical model establishment. As a result of this study, the highlight is found as the use of causally-related variables in the causality-based GEE models is much more appropriate than in the non-causality based GEE models for determining the economic growth profiles of these countries.

ilustraciones, fotografías a color, gráficas Dada una nueva tipología de fraude, en la que cajeros automáticos son bloqueados con el propósito de crear una distracción de los usuario; lo cual, permite el cambio de una tarjeta asociada a un producto financiero y la captura de su clave. El presente trabajo, propone un modelo lineal generalizado como herramienta de pronóstico de la probabilidad de ocurrencia de fraude, mediante la estructuración de una base de datos, extraída de vídeos por medio de redes neuronales convolucionales. Estos modelos estiman la presencia de personas, encontrando el punto de partida para realizar el rastreo de cada individuo por medio de redes neuronales siamesas. La metodología, permite la construcción de covariables en función de la ubicación espacio-temporal de las personas en el lugar de los hechos, insumo que permite la identificación del modelo lineal. (Texto tomado de la fuente) Given a new type of fraud, in which ATMs are blocked with the purpose of creating a distraction for users that allows them to change a card associated with a financial product and capture the password. This paper proposes a Generalized Linear Model as a tool for forecasting the probability of fraud occurrences, by structuring a database extracted from videos by means of Convolutional Neural Networks. This model estimate the presence of people, by finding the starting point to track each individual with a Siamese Neural Networks. The proposal enable the construction of covariates based on the spatio-temporal location of the people at the scene of the events. Input that allows the identification of the lineal model. Maestría Magíster en Ciencias - Estadística

The aim is to describe the Typhus group (TG) Rickettsia infection in dogs and to identify factors associated with this infection. We collected blood samples and gathered exposure and clinical data of 142 dogs from a rural community of Yucatan. The Rickettsia group was determined by semi-nested PCR. Generalized linear models with binomial error distribution were used to model the associated factors from the dog sample for risk ratio (RR) estimation. Thirty-four dogs (23.9%) showed molecular evidence of TG Rickettsia DNA. The multivariate model showed that mixed-breed dogs (RR = 0.06) and dogs that had received antiparasitic treatment (RR = 0.049) had a lower risk of getting infected, taking as reference the purebred group and the non-treated dogs, respectively. Looking at variable interactions, adult dogs without outdoor activities had a lower infection risk than puppies (RR = 0.26). Among dogs with antiparasitic treatment, females had a higher infection risk than male dogs (RR = 26.2). The results showed enzootic TG Rickettsia circulation in dogs of a rural community. The factors outdoor activities, age and previous antiparasitic treatment, as well as the clinical variables signs of hemorrhages and epistaxis, were associated with a less chance of natural infection in the studied dogs. Prevention and control of the enzootic transmission risk of TG Rickettsia should help to reduce the potential zoonotic transmission of this pathogen.

These R scripts contain the code to replicate the analyses performed in Demographic consequences of changing environmental periodicity, Ecology. Vital-rate estimation We used the demographic data of each species to model periodic differences in vital rates for each life-cycle stage using generalized linear models and mixed models (GLMs for the deterministic vital rates in dewy pines and GLMMs for all other vital rates). More specifically, we tested for the effect of season on marmot and meerkat vital rates and of post-fire habitat states (time since fire, TSF) on dewy pine vital rates. For all three species, we estimated stage-specific survival and probability of transition to another stage using a binomial distribution, and reproductive output using a Poisson distribution. In addition, we estimated meerkat helper emigration probability and dewy-pine flowering probability using a binomial distribution. In addition to fixed effects, we incorporated year random effects in all appropriate models to model year-specific differences among season-specific average vital-rates (for marmots and meerkats) or among vital-rate averages (for dewy pines). For the meerkats and dewy pines, we also incorporated the fixed effect of density on vital rates. For each modeled vital rate, we first selected the best random-effect structure where appropriate (i.e., testing whether a random effect on the average vital rate and the slope between seasons outperformed a random effect on the average vital rate only). We then selected the best fixed-effect variables using the Akaike Information Criterion corrected for small sample size (AICc). When the difference in AICc values was not significant (dAICcPopulation projections We used the predictions of the vital-rate models described above to parameterize periodic matrix population models (MPMs) and project the population dynamics for 100 years according to two scenarios—control and perturbed—to assess the effect of changes in vital-rate periodicity on the stochastic growth rate (log λS), the variance of 100 annual growth rates (var(log λ)), and the probability of quasi-extinction (pqext, i.e., the ratio of simulations leading to quasi-extinction out of 500 simulations, with a threshold set at 15% of the minimum observed abundance or number of reproductive individuals for the marmot and meerkat populations, and at 50% of the minimum observed aboveground or seed bank abundance for the dewy pines). Yellow-bellied marmot and meerkat (seasonal dynamics) For the meerkats and marmots, we projected population dynamics under changing patterns of vital rates for which the year random effect in the GLMM was applied both on the mean vital-rate estimate (i.e., the model intercept) and the difference between seasons. That is, for the marmots, we assessed the population responses to changes in the seasonal patterns of yearling and non-reproductive and reproductive adult survival. For the meerkats, we did so for subadult, helper, and dominant survival, helper emigration, transition from helper to dominant, and helper and dominant recruitment. We used the above models to predict season- and year-specific vital rates and parameterize periodic matrix population models (MPMs) for each season. We then simulated population dynamics by projecting MPMs representing low- (LS) and high-seasonality (HS) years for each of the vital rates. Half of all years in which a vital rate was estimated were considered as LS and the other half as HS. That is, for a given vital rate, we defined the threshold between LS and HS years as the 50th percentile of all year-specific absolute differences between seasons. In addition, we used a control scenario, where we projected the population dynamics using MPMs representing both high- and low-seasonality years indistinctly. We performed 500 simulations, each starting with the same population vector, and projected the population dynamics for 100 years. For each simulation, we randomly selected 100 years corresponding to each scenario (LS, HS, or control). In each step of the simulation, all vital rates were predicted based on the same randomly selected year. This allowed us to maintain within-year vital-rate correlation. The predicted vital rates were then used to build the corresponding period-specific MPM. Dewy pine (Multi-year habitat succession and periodicity in fire regimes) For dewy pines, we simulated two distinct types of fire regimes, each with two frequencies: (1) periodic burning occurring systematically every 15 or 30 years, and (2) stochastic fires occurring on average every 15 or 30 years. In all simulations, dewy pines transition deterministically through the first four post-fire states, TSF0 to TSF3. Under periodic fires, once in the fifth state, TSF>3, the population remains in that state until the next fire (15 or 30 years after the previous fire) and then transitions to TSF0. Under stochastic fires, the population can transition from TSF>3 to TSF0 conditional on fire frequency (p). This is done using a Markov-chain approach. In addition to the different periodic patterns in fire regimes, we perturbed periodic patterns in habitat succession in natural heathlands by introducing an additional human disturbance (i.e., using vital rates from populations under a browsing-induced disturbance) first only in the years of the last post-fire state (TSF>3), and then increasingly in the previous states (i.e., in TSF3 and >3, in TSF2, 3, and >3, etc.) until all post-fire states of a natural population were perturbed. We used TSF-specific vital-rate predictions from our models and rates on seed-bank dynamics to build MPMs for each TSF. We used these MPMs to project population dynamics for 100 years using 500 simulations, in each iteration randomly sampling among MPMs describing one of the two natural populations. In addition, for each iteration in the stochastic post-fire state TSF>3, we randomly sampled a year-specific MPM. We increased browsing pressure by replacing MPMs associated with natural habitat conditions by MPMs parameterized with vital rates estimated from human-disturbed populations, for any given TSF. We then compared scenarios of increasing anthropogenic pressures to the control scenario (i.e., natural populations with no browsing perturbation) in each fire regime. Population responses to vital-rate periodic patterns under density dependence For the meerkats and dewy pines, projections of population dynamics incorporated density dependence. That is, at each iteration of the simulations, population density was estimated and used to predict vital rates and parameterize an MPM from these predictions at the next iteration. We then compared density-dependent projections to ones where the density input during parameter estimation was fixed at constant average values obtained from the observed data. Analysis of the simulations results For all three systems and for each simulation, we recorded the stochastic growth rate log λS. We also investigated the effect of changes in vital-rate patterns on the variance in 100 annual log λ, var(log λ), and the quasi-extinction probability pqext. We checked the overlap of the distributions of each metric (i.e., the mean and the 2.5th and 97.5th percentiles) across the 500 simulations between two scenarios. We considered a metric to differ between scenarios when 95% of the distributions (i.e., between the 2.5th and 97.5th percentiles) did not overlap. Comparing population sensitivity to changes in periodic environmental patterns vs. stochastic environmental variation We compared our results from periodic population models to “classic” assessments of population fitness sensitivity to stochastic environmental variation. To do so, for marmots and meerkats, we computed the stochastic elasticities of the population growth rate to changes in the mean and standard deviation of vital rates. We then calculated the relative importance of the stochastic elasticity of the growth rate due to changes in the variability of vital rates compared to changes in their mean. For dewy pines, as the effects of periodic patterns consisted of changing the sequences of post-fire habitat states, we explored the link between the effects of human-induced disturbances in various post-fire habitat states and the role of these states in shaping population dynamics. We thus used the megamatrix approach to calculate the elasticity of the population growth rate to each post-fire habitat state., Additional funding acknowledgment: University of Pretoria MAVA Foundation Ministerio de Economía y Competitividad (CGL2011-28759/BOS) Ministerio de Economía y Competitividad (CGL2015-64007-P) National Geographic Society (8140-06) National Science Foundation (DBI-1226713) National Science Foundation (DEB-1557130)

We propose a simple method for evaluating the model that has been chosen by an adaptive regression procedure, our main focus being the lasso. This procedure deletes each chosen predictor and refits the lasso to get a set of models that are "close" to the chosen "base model," and compares the error rates of the base model with that of nearby models. If the deletion of a predictor leads to significant deterioration in the model's predictive power, the predictor is called indispensable; otherwise, the nearby model is called acceptable and can serve as a good alternative to the base model. This provides both an assessment of the predictive contribution of each variable and a set of alternative models that may be used in place of the chosen model. We call this procedure "Next-Door analysis" since it examines models "next" to the base model. It can be applied to supervised learning problems withLes auteurs proposent une méthode simple pour l’évaluation de modèles choisis par une procédure de régression adaptative telle que le lasso, sur lequel ils se concentrent. Leur procédure consiste à retirer chaque prédicteur à tour de rôle et à réajuster le lasso afin d’obtenir un ensemble de modèles qui sont près du modèle de base. Ils comparent ensuite le taux d’erreur du modèle de base avec ceux de son voisinage. Lorsque le retrait d’une variable conduit à une baisse marquée de la puissance prédictive du modèle, le prédicteur est considéré comme indispensable. Sinon, il est jugé acceptable et peut servir de remplaçant pour le modèle de base. Cette approche permet à la fois de mesurer la contribution prédictive de chaque variable et de constituer un ensemble de modèles de remplacement. Les auteurs ont baptisé cette approche « l’analyse de la porte voisine » puisqu’elle consiste à examiner des modèles près du modèle de base. Celle-ci peut être appliquée aux problèmes d’apprentissage supervisé avec une pénalisation

In the context of generalized linear models, most of the recent studies were on logistic regression models and many of them focussed on optimal experimental designs with concentration on D-optimality. In this research, two- and three-variable Poisson regression models were considered for E-optimization on restricted design space [0, 1]. The two-variable Poisson regression model was not optimal at 3-design points, but was found to be E-optimal at 4-design points (1, 1), (0, 0), (0, 1) and (1, 0) with equal design weights of 0.25. The three-variable Poisson regression model was E-optimal at 4-design points (0, 0, 1), (0, 1, 0), (1, 1, 1) and (1, 0, 0) with each design point having design weights of 0.25. The prediction error variance (PEV) for the two-variable Poisson regression model is 0.35 and that of the three-variable Poisson regression model is 0.68. From this research, the two-variable Poisson regression model is preferred to the three-variable Poisson regression model because of smaller PEV. Keywords: E-optimality; Fisher Information Matrix; Poisson Regression Model; Prediction Error Variance

Analyzing how the proximity to certain features or particular places in a city increases or decreases crime risk across space is a fundamental issue in quantitative criminology from both explanatory and predictive perspectives. Regarding the latter, the detection of high-risk cells is of special interest for practical reasons. There are several statistical modeling approaches that can be implemented in order to fulfil these two main objectives. The purpose of this study is to compare risk terrain modeling (RTM), a method widely used among quantitative criminologists, with a non-linear effects model that considers a non-linear function of distances to the selected places. To this end, a dataset containing crime events recorded in Valencia (Spain) along four years was used to perform the comparison. Several socio-demographic covariates and a selection of places in the city were considered for modeling crime counts with both the RTM and the non-linear approaches. The two modeling techniques were moderately coherent with regard to detecting certain types of places as responsible of higher (or lower) crime rates, but several differences arose. Furthermore, the non-linear model was more accurate than RTM to predict future crime occurrences for each of the three crime types that were considered for the analysis. In conclusion, the possibility of modeling the effect of a place on crime risk through non-linear functions appears as one competitive alternative or at least complement to RTM that may deserve further consideration.

This article concerns maximum penalized likelihood estimation in misspecified generalized linear models with independent and identically distributed observations. A new method for simultaneous mode...

The inference procedure under generalized linear model breaks down when there exists too many zeros in count data that the parent distribution cannot accommodate. We argue that the process of separ...

We consider the problem of fitting a generalized linear model with a three-dimensional image covariate, such as one obtained by functional magnetic resonance imaging (fMRI). A major challenge for f...

Bayesian approaches for criterion based selection include the marginal likelihood based highest posterior model (HPM) and the deviance information criterion (DIC). The DIC is popular in practice as it can often be estimated from sampling-based methods with relative ease and DIC is readily available in various Bayesian software. We find that sensitivity of DIC-based selection can be high, in the range of 90–100%. However, correct selection by DIC can be in the range of 0–2%. These performances persist consistently with increase in sample size. We establish that both marginal likelihood and DIC asymptotically disfavour under-fitted models, explaining the high sensitivities of both criteria. However, mis-selection probability of DIC remains bounded below by a positive constant in linear models with g-priors whereas mis-selection probability by marginal likelihood converges to 0 under certain conditions. A consequence of our results is that not only the DIC cannot asymptotically differentiate between the data-generating and an over-fitted model, but, in fact, it cannot asymptotically differentiate between two over-fitted models as well. We illustrate these results in multiple simulation studies and in a biomarker selection problem on cancer cachexia of non-small cell lung cancer patients. We further study the performances of HPM and DIC in generalized linear model as practitioners often choose to use DIC that is readily available in software in such non-conjugate settings.

We explore the possibility of modeling clustered count data using the Poisson Inverse Gaussian distribution. We develop a regression model, which relates the number of mastitis cases in a sample of dairy farms in Ontario, Canada, to various farm level covariates, to illustrate the methodology. Residual plots are constructed to explore the quality of the fit. We compare the results with a negative binomial regression model using maximum likelihood estimation, and to the generalized linear mixed regression model fitted in SAS.

It is not generally feasible to pick any given marginal distribution and assume there will be a way to apply a link function to add fixed and random effects in a spatial generalized linear model. We introduce an adjustment to spatial copula processes called a non-Gaussian translation process that will allow for the specification of any marginal distribution with a closed-form density function in a single unified framework. While translation processes do not preserve the exact marginal structure, they allow for fixed effects to be included in a non-Gaussian spatial model without needing to define a link function, as well as providing a number of other computational and modeling benefits. Non-Gaussian translation processes are compared theoretically and via simulation with traditional link function approaches and spatial copula processes and are shown to perform similarly in cases where all three models can effectively be used. A daily precipitation data set is analyzed with elevation as a predictor variable that includes a majority of observations being 0. Out-of-sample predictions are evaluated, and it is determined that the model is effective when compared to a two-stage prediction model and a Bayesian power truncated normal model.

The generalized Poisson regression model has been used to model dispersed count data. It is a good competitor to the negative binomial re- gression model when the count data is over-dispersed. Zero-inflated Poisson and zero-inflated negative binomial regression models have been proposed for the situations where the data generating process results into too many zeros. In this paper, we propose a zero-inflated generalized Poisson (ZIGP) regression model to model domestic violence data with too many zeros. Es- timation of the model parameters using the method of maximum likelihood is provided. A score test is presented to test whether the number of zeros is too large for the generalized Poisson model to adequately fit the domestic violence data.