GWRM: An R Package for Identifying Sources of Variation in Overdispersed Count Data.
Ontology highlight
ABSTRACT:
SUBMITTER: Vilchez-Lopez S
PROVIDER: S-EPMC5148598 | biostudies-literature | 2016
REPOSITORIES: biostudies-literature
Publications
GWRM: An R Package for Identifying Sources of Variation in Overdispersed Count Data.
Vílchez-López Silverio S Sáez-Castillo Antonio José AJ Olmo-Jiménez María José MJ
PloS one 20161209 12
Understanding why a random variable is actually random has been in the core of Statistics from its beginnings. The generalized Waring regression model for count data explains that inherent variability is given by three possible sources: randomness, liability and proneness. The model extends the negative binomial regression model and it is not included in the family of generalized linear models. In order to avoid that shortcoming, we developed the GWRM R package for fitting, describing and valida ...[more]
Similar Datasets
Project description:A generic random effects formulation for the Dirichlet negative multinomial distribution is developed together with a convenient regression parameterization. A simulation study indicates that, even when somewhat misspecified, regression models based on the Dirichlet negative multinomial distribution have smaller median absolute error than generalized estimating equations, with a particularly pronounced improvement when correlation between observations in a cluster is high. Estimation of explanatory variable effects and sources of variation is illustrated for a study of clinical trial recruitment.
| S-EPMC3590929 | biostudies-literature
Project description:Count datasets are traditionally analyzed using the ordinary Poisson distribution. However, said model has its applicability limited, as it can be somewhat restrictive to handling specific data structures. In this case, the need arises for obtaining alternative models that accommodate, for example, overdispersion and zero modification (inflation/deflation at the frequency of zeros). In practical terms, these are the most prevalent structures ruling the nature of discrete phenomena nowadays. Hence, this paper's primary goal was to jointly address these issues by deriving a fixed-effects regression model based on the hurdle version of the Poisson-Sujatha distribution. In this framework, the zero modification is incorporated by considering that a binary probability model determines which outcomes are zero-valued, and a zero-truncated process is responsible for generating positive observations. Posterior inferences for the model parameters were obtained from a fully Bayesian approach based on the g-prior method. Intensive Monte Carlo simulation studies were performed to assess the Bayesian estimators' empirical properties, and the obtained results have been discussed. The proposed model was considered for analyzing a real dataset, and its competitiveness regarding some well-established fixed-effects models for count data was evaluated. A sensitivity analysis to detect observations that may impact parameter estimates was performed based on standard divergence measures. The Bayesian p-value and the randomized quantile residuals were considered for the task of model validation.
| S-EPMC8224290 | biostudies-literature
Project description:IntroductionCount outcomes in tobacco research are often analyzed with the Poisson distribution. However, they often exhibit features such as overdispersion (variance larger than expected) and zero inflation (extra zeros) that violate model assumptions. Furthermore, longitudinal studies have repeated measures that generate correlated counts. Failure to account for overdispersion, zero inflation, and correlation can yield incorrect statistical inferences. Thus, it is important to familiarize researchers with proper models for such data.Aims and methodsPoisson and Negative Binomial models with correlated random effects with and without zero inflation are presented. The illustrative data comes from a study comparing a mindfulness training app (Craving to Quit [C2Q], n = 60) with a control app (experience sampling-only app, n = 66) on smoking frequency at 1, 3, and 6 months. Predictors include app, time, the app-by-time interaction, and baseline smoking. Each model is evaluated in terms of accounting for zero inflation, overdispersion, and correlation in the data. Emphasis is placed on evaluating model fit, subject-specific interpretation of effects, and choosing an appropriate model.ResultsThe hurdle Poisson model provided the best fit to the data. Smoking abstinence rates were 33%, 32%, and 28% at 1, 3, and 6 months, respectively, with variance larger than expected by a factor >7 at each follow-up. Individuals on C2Q were less likely to achieve abstinence across time but likely to smoke fewer cigarettes if smoking.ConclusionsThe models presented are specifically suited for analyzing correlated count outcomes and account for zero inflation and overdispersion. We provide guidance to researchers on the use of these models to better inform nicotine and tobacco research.ImplicationsIn tobacco research, count outcomes are often measured repeatedly on the same subject and thus correlated. Such outcomes often have many zeros and exhibit large variances relative to the mean. Analyzing such data require models specifically suited for correlated counts. The presented models and guidelines could improve the rigor of the analysis of correlated count data and thus increase the impact of studies in nicotine and tobacco research using such outcomes.
| S-EPMC10077942 | biostudies-literature
Project description:Data collected in various scientific fields are count data. One way to analyze such data is to compare the individual levels of the factor treatment using multiple comparisons. However, the measured individuals are often clustered - e.g. according to litter or rearing. This must be considered when estimating the parameters by a repeated measurement model. In addition, ignoring the overdispersion to which count data is prone leads to an increase of the type one error rate. We carry out simulation studies using several different data settings and compare different multiple contrast tests with parameter estimates from generalized estimation equations and generalized linear mixed models in order to observe coverage and rejection probabilities. We generate overdispersed, clustered count data in small samples as can be observed in many biological settings. We have found that the generalized estimation equations outperform generalized linear mixed models if the variance-sandwich estimator is correctly specified. Furthermore, generalized linear mixed models show problems with the convergence rate under certain data settings, but there are model implementations with lower implications exists. Finally, we use an example of genetic data to demonstrate the application of the multiple contrast test and the problems of ignoring strong overdispersion.
| S-EPMC9042126 | biostudies-literature
Project description:Aggregation of macroparasites among hosts is a near-universal pattern, and has important consequences for the stability of host-parasite associations and the impacts of disease. Identifying which potential drivers are contributing to levels of aggregation observed in parasite-host associations is challenging, particularly for observational studies. We apply beta regressions in a Bayesian framework to determine predictors of aggregation, quantified using Poulin's index of discrepancy (D), for 13 species of parasites infecting Icelandic Rock Ptarmigan (Lagopus muta) collected over 12 years. 1,140 ptarmigan were collected using sampling protocols maximizing consistency of sample sizes and of composition of host ages and sexes represented across years from 2006-2017. Parasite species, taxonomic group (insect, mite, coccidian, or nematode), and whether the parasite was an ecto- or endoparasite were tested as predictors of aggregation, either alone or by modulating an effect of parasite mean abundance on D. Parasite species was an important predictor of aggregation in models. Despite variation in D across samples and years, relatively consistent aggregation was demonstrated for each specific host-parasite association, but not for broader taxonomic groups, after taking sample mean abundance into account. Furthermore, sample mean abundance was consistently and inversely related to aggregation among the nine ectoparasites, however no relationship between mean abundance and aggregation was observed among the four endoparasites. We discuss sources of variation in observed aggregation, sources both statistical and biological in nature, and show that aggregation is predictable, and distinguishable, among infecting species. We propose explanations for observed patterns and call for the review and re-analysis of parasite and other symbiont distributions using beta regression to identify important drivers of aggregation-both broad and association-specific.
| S-EPMC9419717 | biostudies-literature
Project description:A new software package for the Julia language, CountTimeSeries.jl, is under review, which provides likelihood based methods for integer-valued time series. The package's functionalities are showcased in a simulation study on finite sample properties of Maximum Likelihood (ML) estimation and three real-life data applications. First, the number of newly infected COVID-19 patients is predicted. Then, previous findings on the need for overdispersion and zero inflation are reviewed in an application on animal submissions in New Zealand. Further, information criteria are used for model selection to investigate patterns in corporate insolvencies in Rhineland-Palatinate. Theoretical background and implementation details are described, and complete code for all applications is provided online. The CountTimeSeries package is available at the general Julia package registry.
| S-EPMC8228825 | biostudies-literature
Project description:Count data are common endpoints in clinical trials, for example magnetic resonance imaging lesion counts in multiple sclerosis. They often exhibit high levels of overdispersion, that is variances are larger than the means. Inference is regularly based on negative binomial regression along with maximum-likelihood estimators. Although this approach can account for heterogeneity it postulates a common overdispersion parameter across groups. Such parametric assumptions are usually difficult to verify, especially in small trials. Therefore, novel procedures that are based on asymptotic results for newly developed rate and variance estimators are proposed in a general framework. Moreover, in case of small samples the procedures are carried out using permutation techniques. Here, the usual assumption of exchangeability under the null hypothesis is not met due to varying follow-up times and unequal overdispersion parameters. This problem is solved by the use of studentized permutations leading to valid inference methods for situations with (i) varying follow-up times, (ii) different overdispersion parameters, and (iii) small sample sizes.
| S-EPMC6587510 | biostudies-literature
Project description:BackgroundMethylation measures quantified by microarray techniques can be affected by systematic variation due to the technical processing of samples, which may compromise the accuracy of the measurement process and contribute to bias the estimate of the association under investigation. The quantification of the contribution of the systematic source of variation is challenging in datasets characterized by hundreds of thousands of features.In this study, we introduce a method previously developed for the analysis of metabolomics data to evaluate the performance of existing normalizing techniques to correct for unwanted variation. Illumina Infinium HumanMethylation450K was used to acquire methylation levels in over 421,000 CpG sites for 902 study participants of a case-control study on breast cancer nested within the EPIC cohort. The principal component partial R-square (PC-PR2) analysis was used to identify and quantify the variability attributable to potential systematic sources of variation. Three correcting techniques, namely ComBat, surrogate variables analysis (SVA) and a linear regression model to compute residuals were applied. The impact of each correcting method on the association between smoking status and DNA methylation levels was evaluated, and results were compared with findings from a large meta-analysis.ResultsA sizeable proportion of systematic variability due to variables expressing 'batch' and 'sample position' within 'chip' was identified, with values of the partial R2 statistics equal to 9.5 and 11.4% of total variation, respectively. After application of ComBat or the residuals' methods, the contribution was 1.3 and 0.2%, respectively. The SVA technique resulted in a reduced variability due to 'batch' (1.3%) and 'sample position' (0.6%), and in a diminished variability attributable to 'chip' within a batch (0.9%). After ComBat or the residuals' corrections, a larger number of significant sites (k?=?600 and k?=?427, respectively) were associated to smoking status than the SVA correction (k?=?96).ConclusionsThe three correction methods removed systematic variation in DNA methylation data, as assessed by the PC-PR2, which lent itself as a useful tool to explore variability in large dimension data. SVA produced more conservative findings than ComBat in the association between smoking and DNA methylation.
| S-EPMC5863487 | biostudies-literature
Project description:BackgroundDifferential expression analysis based on "next-generation" sequencing technologies is a fundamental means of studying RNA expression. We recently developed a multi-step normalization method (called TbT) for two-group RNA-seq data with replicates and demonstrated that the statistical methods available in four R packages (edgeR, DESeq, baySeq, and NBPSeq) together with TbT can produce a well-ranked gene list in which true differentially expressed genes (DEGs) are top-ranked and non-DEGs are bottom ranked. However, the advantages of the current TbT method come at the cost of a huge computation time. Moreover, the R packages did not have normalization methods based on such a multi-step strategy.ResultsTCC (an acronym for Tag Count Comparison) is an R package that provides a series of functions for differential expression analysis of tag count data. The package incorporates multi-step normalization methods, whose strategy is to remove potential DEGs before performing the data normalization. The normalization function based on this DEG elimination strategy (DEGES) includes (i) the original TbT method based on DEGES for two-group data with or without replicates, (ii) much faster methods for two-group data with or without replicates, and (iii) methods for multi-group comparison. TCC provides a simple unified interface to perform such analyses with combinations of functions provided by edgeR, DESeq, and baySeq. Additionally, a function for generating simulation data under various conditions and alternative DEGES procedures consisting of functions in the existing packages are provided. Bioinformatics scientists can use TCC to evaluate their methods, and biologists familiar with other R packages can easily learn what is done in TCC.ConclusionDEGES in TCC is essential for accurate normalization of tag count data, especially when up- and down-regulated DEGs in one of the samples are extremely biased in their number. TCC is useful for analyzing tag count data in various scenarios ranging from unbiased to extremely biased differential expression. TCC is available at http://www.iu.a.u-tokyo.ac.jp/~kadota/TCC/ and will appear in Bioconductor (http://bioconductor.org/) from ver. 2.13.
| S-EPMC3716788 | biostudies-literature
Project description:Accurate estimates of animal abundance are essential for guiding effective management, and poor survey data can produce misleading inferences. Aerial surveys are an efficient survey platform, capable of collecting wildlife data across large spatial extents in short timeframes. However, these surveys can yield unreliable data if not carefully executed. Despite a long history of aerial survey use in ecological research, problems common to aerial surveys have not yet been adequately resolved. Through an extensive review of the aerial survey literature over the last 50 years, we evaluated how common problems encountered in the data (including nondetection, counting error, and species misidentification) can manifest, the potential difficulties conferred, and the history of how these challenges have been addressed. Additionally, we used a double-observer case study focused on waterbird data collected via aerial surveys and an online group (flock) counting quiz to explore the potential extent of each challenge and possible resolutions. We found that nearly three quarters of the aerial survey methodology literature focused on accounting for nondetection errors, while issues of counting error and misidentification were less commonly addressed. Through our case study, we demonstrated how these challenges can prove problematic by detailing the extent and magnitude of potential errors. Using our online quiz, we showed that aerial observers typically undercount group size and that the magnitude of counting errors increases with group size. Our results illustrate how each issue can act to bias inferences, highlighting the importance of considering individual methods for mitigating potential problems separately during survey design and analysis. We synthesized the information gained from our analyses to evaluate strategies for overcoming the challenges of using aerial survey data to estimate wildlife abundance, such as digital data collection methods, pooling species records by family, and ordinal modeling using binned data. Recognizing conditions that can lead to data collection errors and having reasonable solutions for addressing errors can allow researchers to allocate resources effectively to mitigate the most significant challenges for obtaining reliable aerial survey data.
| S-EPMC8931709 | biostudies-literature