Count Models Research Articles

Integrated distance sampling models for simple point counts.

Point counts (PCs) are widely used in biodiversity surveys but, despite numerous advantages, simple PCs suffer from several problems: detectability, and therefore abundance, is unknown; systematic spatiotemporal variation in detectability yields biased inferences, and unknown survey area prevents formal density estimation and scaling-up to the landscape level. We introduce integrated distance sampling (IDS) models that combine distance sampling (DS) with simple PC or detection/nondetection (DND) data to capitalize on the strengths and mitigate the weaknesses of each data type. Key to IDS models is the view of simple PC and DND data as aggregations of latent DS surveys that observe the same underlying density process. This enables the estimation of separate detection functions, along with distinct covariate effects, for all data types. Additional information from repeat or time-removal surveys, or variable survey duration, enables the separate estimation of the availability and perceptibility components of detectability with DS and PC data. IDS models reconcile spatial and temporal mismatches among data sets and solve the above-mentioned problems of simple PC and DND data. To fit IDS models, we provide JAGS code and the new "IDS()" function in the R package unmarked. Extant citizen-science data generally lack the information necessary to adjust for detection biases, but IDS models address this shortcoming, thus greatly extending the utility and reach of these data. In addition, they enable formal density estimation in hybrid designs, which efficiently combine DS with distance-free, point-based PC or DND surveys. We believe that IDS models have considerable scope in ecology, management, and monitoring.

Interpretable zero-inflated neural network models for predicting admission counts

Abstract In this paper, we construct interpretable zero-inflated neural network models for modeling hospital admission counts related to respiratory diseases among a health-insured population and their dependants in the United States. In particular, we exemplify our approach by considering the zero-inflated Poisson neural network (ZIPNN), and we follow the combined actuarial neural network (CANN) approach for developing zero-inflated combined actuarial neural network (ZIPCANN) models for modeling admission rates, which can accommodate the excess zero nature of admission counts data. Furthermore, we adopt the LocalGLMnet approach (Richman & Wüthrich (2023). Scandinavian Actuarial Journal, 2023(1), 71–95.) for interpreting the ZIPNN model results. This facilitates the analysis of the impact of a number of socio-demographic factors on the admission rates related to respiratory disease while benefiting from an improved predictive performance. The real-life utility of the methodologies developed as part of this work lies in the fact that they facilitate accurate rate setting, in addition to offering the potential to inform health interventions.

Determinants of confrontation naming deficits on the Boston Naming Test associated with transactive response DNA-binding protein 43 pathology.

To determine whether poorer performance on the Boston Naming Test (BNT) in individuals with transactive response DNA-binding protein 43 pathology (TDP-43+) is due to greater loss of word knowledge compared to retrieval-based deficits. Retrospective clinical-pathologic study of 282 participants with Alzheimer's disease neuropathologic changes (ADNC) and known TDP-43 status. We evaluated item-level performance on the 60-item BNT for first and last available assessment. We fit cross-sectional negative binomial count models that assessed total number of incorrect items, number correct of responses with phonemic cue (reflecting retrieval difficulties), and number of "I don't know" (IDK) responses (suggestive of loss of word knowledge) at both assessments. Models included TDP-43 status and adjusted for sex, age, education, years from test to death, and ADNC severity. Models that evaluated the last assessment adjusted for number of prior BNT exposures. 43% were TDP-43+. The TDP-43+ group had worse performance on BNT total score at first (p = .01) and last assessments (p = .01). At first assessment, TDP-43+ individuals had an estimated 29% (CI: 7%-56%) higher mean number of incorrect items after adjusting for covariates, and a 51% (CI: 15%-98%) higher number of IDK responses compared to TDP-43-. At last assessment, compared to TDP-43-, the TDP-43+ group on average missed 31% (CI: 6%-62%; p = .01) more items and had 33% more IDK responses (CI: 1% fewer to 78% more; p = .06). An important component of poorer performance on the BNT in participants who are TDP-43+ is having loss of word knowledge versus retrieval difficulties.

A Fixed-Point Approach to Unified Prompt-Based Counting

Existing class-agnostic counting models typically rely on a single type of prompt, e.g., box annotations. This paper aims to establish a comprehensive prompt-based counting framework capable of generating density maps for concerned objects indicated by various prompt types, such as box, point, and text. To achieve this goal, we begin by converting prompts from different modalities into prompt masks without requiring training. These masks are then integrated into a class-agnostic counting methodology for predicting density maps. Furthermore, we introduce a fixed-point inference along with an associated loss function to improve counting accuracy, all without introducing new parameters. The effectiveness of this method is substantiated both theoretically and experimentally. Additionally, a contrastive training scheme is implemented to mitigate dataset bias inherent in current class-agnostic counting datasets, a strategy whose effectiveness is confirmed by our ablation study. Our model excels in prominent class-agnostic datasets and exhibits superior performance in cross-dataset adaptation tasks.

A mRNA panel for differentiation between acute exacerbation or pneumonia in COPD patients.

Patients suffering from chronic obstructive pulmonary disease (COPD) are prone to acute exacerbations (AECOPD) or community acquired pneumonia (CAP), both posing severe risk of morbidity and mortality. There is no available biomarker that correctly separates AECOPD from COPD. However, because CAP and AECOPD differ in aetiology, treatment and prognosis, their discrimination would be important. This study analysed the ability of selected candidate transcripts from peripheral blood mononuclear cells (PBMCs) to differentiate between patients with AECOPD, COPD & CAP, and CAP without pre-existing COPD. In a previous study, we identified differentially regulated genes between CAP and AECOPD in PBMCs. In the present new cohort, we tested the potential of selected candidate PBMC transcripts to differentiate at early time points AECOPD, CAP+COPD, and CAP without pre-existing COPD. Expression of YWHAG, E2F1 and TDRD9 held predictive power: This gene set predicted diseases markedly better (model accuracy up to 100%) than classical clinical markers like CRP, lymphocyte count and neutrophil count (model accuracy up to 82%). In summary, in our cohort expression levels of YWHAG, E2F1 and TDRD9 differentiated with high accuracy between COPD patients suffering from acute exacerbation or CAP.

SCS-YOLOv5s: A cattle detection and counting method for complex breeding environment

Cattle detection and counting is one of the most important topics in the development of modern agriculture and animal husbandry. The traditional manual monitoring methods are inefficient and constrained by factors such as site. To solve the above problems, a SCS-YOLOv5 cattle detection and counting model for complex breeding scenarios is proposed. The original SPPF module is replaced in the YOLOv5 backbone network with a CSP structured SPPFCSPC. A CA (Coordinate Attention) mechanism is added to the neck network, as well as the SC (Standard Convolution) of the Neck network is replaced with a light convolution GSConv and Slim Neck is introduced, and training strategies such as multi-scale training are also employed. The experimental results show that the proposed method enhances the feature extraction ability and feature fusion ability, balances the localization accuracy and detection speed, and improves the use effect in real farming scenarios. The Precision of the improved network model is improved from 93.2% to 95.5%, mAP@0.5 is improved from 94.5% to 95.2%, the RMSE is reduced by about 0.03, and the FPS reaches 88. Compared with other mainstream algorithms, the comprehensive performance of SCS-YOLOv5 s is in a leading position, with fewer missed and false detections, and the strong robustness and generalization ability of this model are proved on multi-category public datasets. Applying the improvement ideas in this paper to YOLOv8 s also yields an increase in accuracy. The improved method in this study can greatly improve the accuracy of cattle detection and counting in complex environments, and has good real-time performance, so as to provide technical support for large-scale cattle breeding.

A rapid, low-cost wheat spike grain segmentation and counting system based on deep learning and image processing

Spike grain number is a crucial parameter when estimating wheat yield. However, most methods predominantly focus on a single period, lacking universality. To achieve rapid and intelligent wheat spike grain counting from the filling stage to the mature stage, this study used different smartphones to obtain wheat ear images of different varieties and constructs the spike grain dataset required for deep learning segmentation model through image normalization and data enhancement. Then we propose a segmentation model called Multi-Attention TransUNet (MATransUNet) and a spike grain counting model (SGCountM) to achieve wheat spike grain counting. Our results indicate that, comparing with other models, MATransUNet achieves the best segmentation performance with strong robustness and generalization capabilities, achieving a mean intersection over union(mIoU) of 85.93%. Through the comparative analysis with manual counting results, our Method I, doubling the number of grains on one side of the ears, results in a root mean squared error (RMSE) of 4.38 and a coefficient of determination(R2) of 0.85; while the Method II, adding the grains on both sides of the ears, yields a RMSE of 3.13 and a R2 of 0.93. The counting accuracy significantly improves compared to not using the SGCountM. Moreover, transfer learning notably enhances the accuracy of the segmentation and counting model, enabling accurate spike grain counting from the filling stage to the mature stage. Finally, we integrate MATransUNet and SGCountM to design and implement a wheat spike grain segmentation and counting system (WeChat mini program). In practical tests, the mini program can effectively, accurately segments and counts wheat spike grain, demonstrating its feasibility and effectiveness. This research provides valuable insights into the intelligent application of wheat spike grain counting.

A new over-dispersed count model based on Poisson-Geometric convolution

A new two-parameter discrete distribution, namely the PoiG distribution, is derived by the convolution of a Poisson variate and an independently distributed geometric random variable. This distribution generalizes both the Poisson and geometric distributions and can be used for modeling over-dispersed as well as equi-dispersed count data. A number of important statistical properties of the proposed count model, such as the probability generating function, the moment generating function, the moments, the survival function and the hazard rate function have been studied in detail. Monotonic properties such as the log concavity are also studied. Stochastic ordering for the proposed model is also investigated in detail. Method of moments and the maximum-likelihood estimators of the parameters of the proposed model are found to work well in simulation experiments as well as in the real-life data analysis. Asymptotic confidence intervals and the coverage probabilities of the parameters are also studied through extensive simulation experiments. Two real life modelling applications demonstrate that the proposed distribution may prove to be useful for the practitioners for modelling over-dispersed count data compared to its closest competitors.

The importance of independence in unmarked spatial capture–recapture analysis

Wildlife populations can be unmarked, meaning individuals lack distinguishing features for individual identification. Populations may also exhibit non‐independent movements, meaning individuals move together. For populations of either unmarked or non‐independent individuals, models based on spatial capture–recapture (SCR) approaches can be used to estimate abundance, density, and other parameters critical for monitoring, management, and conservation. However, when individuals are both unmarked and non‐independent, few model options are available. One approach has been to apply unmarked models and not address the non‐independence despite unquantified impacts on bias, precision, and the ability to make robust ecological inferences. We conducted a simulation study to quantify the impact of non‐independence on the performance of spatial count (SC) and spatial partial identity models (SPIM) – two SCR‐based unmarked modeling approaches – and used the performance of fully marked and independent SCR as a reference. We varied the levels of non‐independence (aggregation and cohesion), detection probability, and the number of partial identity covariates used to resolve identities in SPIM estimation. We expected abundance estimates to be increasingly biased and precise as aggregation and cohesion increased. Results showed that models indeed became less robust to increasing non‐independence, but importantly suggested that only SPIM could be reliably applied under low levels of cohesion when sufficient partial identity covariates are available. SC yielded consistently biased estimates with poor precision. SCR was consistently robust across combinations of aggregation and cohesion, as expected. We therefore advise against the use of SC models for estimating population parameters when individuals are known to be non‐independent, caution that SPIM may be used under narrow ecological conditions, and encourage continued investigations into sampling design and methods development for estimating populations of unmarked and non‐independent individuals.

Head circumference at birth and postnatal growth trajectory in vulnerable groups from Argentina.

To investigate the association between the anthropometric status at birth and brain and bone growth during the first year of life. According to the brain-sparing hypothesis, we expect catch-up to be faster in head circumference (HC) than in body length. This is a longitudinal design that included Argentinian infants under 12 months of age with at least three anthropometric records. We classified study participants into four growth status categories according to z-scores for HC (HCZ) and length (LAZ) at birth, with z-score = -2 as a threshold. We used the Count model to describe growth trajectories in HC and length in the first year of life according to the growth status at birth. Recovery indicator for HC and length was taken as the time until the predicted growth trajectory surpassed the threshold curve predicted by z-score = -2 for age. Growth models included 3399 infants. There were significant differences in the growth parameters between groups in all cases (p < 0.05). Within the group with a low HCZ and a low LAZ at birth, HC recovery was faster than length. In the case of a low z-score for only one of the variables, newborns with a low HCZ recovered faster than individuals born with a low LAZ. The postnatal growth pattern in HC and length is associated with the growth status of HC and length at birth. As we hypothesized, the fastest postnatal recovery occurs for HC in cases of intrauterine delayed growth.

Empirical analysis of drought-induced cattle destocking in South Africa.

The study estimated the determinants of smallholder cattle farmers' decision to destock during drought, using a count model and accounted for socioeconomic and farmer-specific factors.

Extensive hypothesis testing for estimation of crash frequency models

Estimating crash data count models poses a significant challenge which requires extensive knowledge, experience, and meticulous hypothesis testing to capture underlying trends. Simultaneous consideration of multiple modelling aspects is required including, among others, functional forms, likely contributing factors, and unobserved heterogeneity. However, model development, frequently affected by time and knowledge, can easily overlook crucial modelling aspects such as identification of likely contributing factors, necessary transformations, and distributional assumptions. To facilitate model development and an estimation that can extract as many insights as possible, an optimization framework is proposed to generate and simultaneously test a diverse array of hypothesis. The framework comprises a mathematical programming formulation and three alternative solution algorithms. The objective function involves minimizing the Bayesian Information Criterion (BIC) to avoid overfitting. The solution algorithms include metaheuristics to deal with an NP-hard problem and search through a complex and nonconvex space. The metaheuristics also enable to handle unique datasets through varying search strategies. The effectiveness of the proposed framework was ascertained using three distinct datasets, and published models used as benchmarks. The results highlighted the ability of the proposed framework to estimate crash data count models, surpassing benchmark models in terms of insights and goodness-of-fit. The framework provides several advantages, such as robust hypothesis testing, uncovering unique specifications and vital insights in the data, and leveraging existing knowledge to enhance search efficiency. The framework also exposes the vulnerability of traditional analyst efforts to fall into local optima, bias, and limitations in creating more efficient models. In a compelling example using crash data from Washington, the proposed framework unveiled insights overlooked by a benchmark published model, identifying speed, interchanges, and grade breaks as likely crash contributors, and revealing the potential danger of excessively wide shoulders. Conversely, the benchmark model identified fewer contributing factors and missed a crucial non-linear relationship between crash safety and shoulder widths. While wider shoulders are typically associated with improved safety, the proposed models suggest a safety threshold beyond which further widening could decrease safety. The introduction of random parameters in the analysis revealed a more nuanced relationship with crash frequency, thereby underlining the limitations of models incapable of capturing heterogeneity.

City on fire: The role of extortion in urban fires

AbstractMost of the literature on fires focuses on the wild or rural environment. Nevertheless, urban fires, and in particular those related to criminal activity, have recently begun to receive greater attention. This study focuses on the analysis of fires in shopping malls, public markets and businesses in Mexico City to study the criminal intention of arson in cities. Using spatial analysis and count models, we study emergency calls (911) from January 2019 to February 2021 to explore the occurrence of fires and extortion in order to predict these events. Our main finding indicates a spatial concentration of fires in the city, with extortion as a significant predictor that increases the occurrence of fires by an average of 16.63%, controlling for non‐intentional factors. In commercial nodes, extortionists may burn down premises that resist extortion in reprisal and as an indirect threat to future victims. The results contribute to the understanding of a new line of research on arson, real estate fraud and financially profitable activities for organised crime. It is possible that ‘professional torches’ are also linked to another, previously unconsidered crime: that of extortion.

Accurate and fast implementation of soybean pod counting and localization from high-resolution image.

Soybean pod count is one of the crucial indicators of soybean yield. Nevertheless, due to the challenges associated with counting pods, such as crowded and uneven pod distribution, existing pod counting models prioritize accuracy over efficiency, which does not meet the requirements for lightweight and real-time tasks. To address this goal, we have designed a deep convolutional network called PodNet. It employs a lightweight encoder and an efficient decoder that effectively decodes both shallow and deep information, alleviating the indirect interactions caused by information loss and degradation between non-adjacent levels. We utilized a high-resolution dataset of soybean pods from field harvesting to evaluate the model's generalization ability. Through experimental comparisons between manual counting and model yield estimation, we confirmed the effectiveness of the PodNet model. The experimental results indicate that PodNet achieves an R2 of 0.95 for the prediction of soybean pod quantities compared to ground truth, with only 2.48M parameters, which is an order of magnitude lower than the current SOTA model YOLO POD, and the FPS is much higher than YOLO POD. Compared to advanced computer vision methods, PodNet significantly enhances efficiency with almost no sacrifice in accuracy. Its lightweight architecture and high FPS make it suitable for real-time applications, providing a new solution for counting and locating dense objects.

Spectral information content of Compton scattering events in silicon photon counting detectors.

Silicon (Si) is a possible sensor material for photon counting detectors (PCDs). A major drawback of Si is that roughly two-thirds of x-ray interactions in the diagnostic energy range are Compton scattering. Because Compton scattering is an energy-insensitive process, it is commonly assumed that Compton events retain little spectral information. To quantify how much information can be recovered from Compton scattering events in models of Si PCDs. We built a simplified model of Si interactions including two interaction mechanisms: photoelectric effect and Compton scattering. We considered three different binning options that represent strategies for handling Compton events: in Compton censoring, all events under 38keV (the maximum energy possible from Compton scattering for a 120keV incident photon) were discarded; in Compton counting, all events between 1 and 38keV were placed into a single bin; in Compton binning, all events were placed into energy bins of uniform width. These were compared to the ideal detector, which always recorded the correct energy (i.e., 100% photoelectric effect). Every photon was assumed to interact once and only once with Si, and the energy bin width was 5keV. In the primary analysis, the Si detector was irradiated with a 120kV spectrum filtered by 30cm of water, with 99.5% of the arriving spectrum above 38keV so that there was good separation between photoelectric effect and Compton scattering, and the figures of merit were the Cramér-Rao lower bound (CRLB) of the variance of iodine and water basis material decomposition images, as well as the CRLB of virtual monoenergetic images (i.e., linear combinations of material images) that maximize iodine CNR or water CNR. We also constructed a local linear estimator that attains the CRLB. In secondary analyses, we applied other sources of spectral distortion: (1) a nonzero minimum energy threshold; (2) coarser, 10keV energy bins; and (3) a model of charge sharing. With our chosen spectrum, 67% of the interactions were Compton scattering. Consistent with this, the material decomposition variance for the Compton censoring model, averaged over both basis materials, was 258% greater than the ideal detector. If Compton events carried no spectral information, the Compton counting model would show similar variance. Instead, its basis material variance was 103% greater than the ideal detector, implying that Compton counts indeed carry significant spectral information. The Compton binning model had a basis material variance 60% greater than the ideal detector. The Compton binning model was not affected by a 5keV minimum energy threshold, but the variance increased from 60% to 107% when charge sharing was included and to 78% with coarser energy bins. For optimized CNR images, the average variance was 149%, 12%, and 10% higher than the ideal detector for the Compton censoring, counting, and binning models, reinforcing the hypothesis that Compton counts are useful for detection tasks and that precise energy assignments are not necessary. Substantial spectral information remains after Compton scattering events in silicon PCDs.

Telematics combined actuarial neural networks for cross-sectional and longitudinal claim count data

AbstractWe present novel cross-sectional and longitudinal claim count models for vehicle insurance built upon the combinedd actuarial neural network (CANN) framework proposed by Wüthrich and Merz. The CANN approach combines a classical actuarial model, such as a generalized linear model, with a neural network. This blending of models results in a two-component model comprising a classical regression model and a neural network part. The CANN model leverages the strengths of both components, providing a solid foundation and interpretability from the classical model while harnessing the flexibility and capacity to capture intricate relationships and interactions offered by the neural network. In our proposed models, we use well-known log-linear claim count regression models for the classical regression part and a multilayer perceptron (MLP) for the neural network part. The MLP part is used to process telematics car driving data given as a vector characterizing the driving behavior of each insured driver. In addition to the Poisson and negative binomial distributions for cross-sectional data, we propose a procedure for training our CANN model with a multivariate negative binomial specification. By doing so, we introduce a longitudinal model that accounts for the dependence between contracts from the same insured. Our results reveal that the CANN models exhibit superior performance compared to log-linear models that rely on manually engineered telematics features.

Count Models of Mortality for Children Under Five Years in Kenya

Count Models of Mortality for Children Under Five Years in Kenya

Parameter-Free and Electron Counting Satisfied Material Representation for Machine Learning Potential Energy and Force Fields.

We proposed a parameter-free volume element representation that satisfies the electron counting model and obtains accurate machine learning potential energy and direct force fitting of randomly perturbed hexagonal BN. Our method preserves permutational, translational, and rotational invariance and can be extended to three-dimensional systems, verified by a system of bulk Si. As a result, we obtained 0.57 meV/atom potential energy root mean squared error (RMSE) and 59 meV/Å force RMSE for perturbed bulk BN systems and 0.43 meV/atom potential energy RMSE and 36 meV/Å force RMSE for perturbed Si systems. In addition, an unbiased perturbation-based data set construction scheme is introduced and a continuous population distribution is obtained with a training data set of 4500, which is about 1 order of magnitude smaller than standard methods based on first-principles molecular dynamics simulations and saves a large amount of computing resources. General validity of our model is verified by structure optimization, molecular dynamics simulations, and extrapolations.

Multilevel Discrete Outcome Modeling for Crash Severity: A Novel Approach for Crash Severity Models

Combining a total crash count model and a crash severity probability model simultaneously to predict crashes by severity is challenging because crash count models are implemented at an aggregate scale, that is, by sites (segments or intersections), while crash severity probability models are at a disaggregate scale, that is, by crashes or persons (drivers or occupants). This study develops a multilevel discrete outcome modeling framework to estimate crash counts by severity for California, U.S., rural four-lane divided highways by estimating and combining three models: a univariate count model for total crashes at the site level predicts total crash counts; an ordered logit discrete outcome crash severity prediction model at the crash-level predicts proportion of crashes by severity; and a proportional regression connection model that predicts proportion of groups of drivers and vehicles to connect the other two models. The combination of all three models provides crash counts by severity at the site level considering both aggregate and disaggregate factors. The prediction from the multilevel discrete outcome model is compared with predictions made by traditional univariate count models by severity for accuracy and transferability using estimation and validation data. Based on the estimation data, the multilevel discrete outcome model predicts injury and fatal crashes more accurately than the univariate count model. For validation data, the multilevel model has slightly better prediction accuracy for minor and possible injuries. Finally, the calibration factor indicates the multilevel discrete outcome model is more transferable than the univariate count model.

Improved Method to Estimating Parameters of a Poisson Hidden Markov Model Using Bayesian Approach

Expectation-Maximization (EM) is a popular method for estimating parameters of the Poisson-Hidden Markov Model (P-HMM). However, over-dispersion in comparison to the Poisson distribution remains a concern. This study developed a Bayesian method to Poisson count models. The study compares the Mean Square Errors and sufficiency of the EM to the Gibbs sampler technique using Akaike Information Criterion, Bayesian Information Criterion, and Deviance Information Criterion as model selectors. The Maximum Likelihood Estimates and Maximum a \textit{posteriori} were calculated using both simulated and real data in this case. The study&amp;#39;s findings indicate that using any of the two approaches depends on the data type and the sample size. Whereas the Poisson Hidden Markov model, which uses the EM algorithm is preferred when using Zero-inflated data with a sample size&amp;nbsp; n &amp;le; 20 , the Bayesian Poisson-Hidden Markov Model, which uses a Gibbs sampler is better used for Heavy and Mixture data types irrespective of the sample size. The model predicted parameters of simulated data with remarkable accuracy and produced some unique statistical property results. This method is applicable to Poisson-hidden Markov models with homogeneous time series.