Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Bayesian modelling”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 343 records · Page 19Linked to original sources
PubMed1987

Suicide risk prediction by computer interview: a prospective study.

A computer interview program that uses a subjective Bayesian probability model to assess suicide risk was evaluated. Predictions made by clinicians for 52 patients were compared with predictions made by the computer for the same patients. The computer was significantly (p = .001) better at predicting attempters, and clinicians were significantly (p = .01) better at predicting nonattempters. An analysis of receiver operating characteristic curves showed that the computer had better overall discrimination, but the difference was nonsignificant.

Decision Making, Computer-Assisted↗
PubMed1984

Clinical inferences and decisions--II. Decision trees, receiver operator curves and subjective probability.

In patient management, clinical decisions follow a logical sequence which can be formally expressed as a decision tree in which the uncertainties associated with each alternative outcome may be made explicit using Bayes' theorem. Where test data is used in the formulation of a decision, the uncertainty associated with the information it conveys may be modified by changing the pass/fail criterion to alter the false positive and false negative error rate. Classical procedures based on information theory are described to illustrate how this may be achieved for any test. When hard data is not available to permit such an approach, the clinician must rely on his own past experience or that of a colleague. Several methods are available for quantifying such experience by estimating subjective probabilities associated with an action or test result. Two simple methods are described for deriving subjective probabilities for subsequent use within a Bayesian decision model.

Bayes Theorem↗
PubMed1981

Case-control diagnosis and Bayesian inference in common viral infections.

The predictive values of symptoms and signs for given diseases are often unknown. The fact that a high proportion of individuals with a certain disease may have a specific group of symptoms (the case-control approach) does not necessarily mean that the specific group of symptoms will allow one reliably to diagnose the disease. This study, utilizing a population based data set for common acute infections, shows that descriptions of common viral illnesses found in medical textbooks that associate illnesses with symptoms do not allow one to predict reliably isolation of the supposed causal organism. Positive predictive value of groups of symptoms for specific viral infections did not exceed 11 percent in this study. However, the data closely fitted the Bayesian statistical model often proposed for such decision making by physicians.

Adolescent↗
PubMed1995

[The prognostic evaluation of acute pancreatitis].

Early identification of severity is one of the most important problems in acute pancreatitis, both for decision-making and classification. Predictive criteria show a wide range of accuracy: clinical examination (on admission: 76-85%); single laboratory data (PCR: 68-98%, C3-C4: 63-72%); multifactorial scoring systems (Ranson: 65-82%, Imrie: 78-95%); diagnostic peritoneal lavage (72-90%); CT features (52-81%). In 1982 we started a prospective evaluation of the prognostic performances of a bayesian statistical model for the prediction of severe vs mild pancreatis and death vs survival, which uses the outcome-related patterns of several variables, assuming their independence, analysed on a data of 44 patients. The performances have been calculated prospectively by comparing the expected vs actual results on 88 further patients (accuracy, sensitivity and specificity, respectively, in the prediction of severe pancreatitis: 92%, 92%, 93%; in the prediction of death: 95%, 97%, 87%). Moreover, the model can represent classes of risk by combining prediction of death + severe pancreatitis (DSP), survival + severe pancreatitis (SSP) and survival + mild pancreatitis (SMP) (accuracy, sensitivity and specificity, respectively, in the prediction of DSP: 97%, 83%, 100%; in the prediction of SSP: 95%, 87%, 97%; in the prediction of SMP: 95%, 97%, 90%). Our model enables clinicians dealing with other population to re-determine different variables or integrate them with new information, whenever available. It seems to be transferable and adaptable, even with a probable further increase of the performances, without compromising the objectivity of the predictive judgement and the homogeneity of the classes of risk.

Acute Disease↗
PubMed1997

Large hierarchical Bayesian analysis of multivariate survival data.

Failure times that are grouped according to shared environments arise commonly in statistical practice. That is, multiple responses may be observed for each of many units. For instance, the units might be patients or centers in a clinical trial setting. Bayesian hierarchical models are appropriate for data analysis in this context. At the first stage of the model, survival times can be modelled via the Cox partial likelihood, using a justification due to Kalbfleisch (1978, Journal of the Royal Statistical Society, Series B 40, 214-221). Thus, questionable parametric assumptions are avoided. Conventional wisdom dictates that it is comparatively safe to make parametric assumptions at subsequent stages. Thus, unit-specific parameters are modelled parametrically. The posterior distribution of parameters given observed data is examined using Markov chain Monte Carlo methods. Specifically, the hybrid Monte Carlo method, as described by Neal (1993a, in Advances in Neural Information Processing 5, 475-482; 1993b, Probabilistic inference using Markov chain Monte Carlo methods), is utilized.

Antineoplastic Combined Chemotherapy Protocols↗
PubMedDate not supplied

Advances in meta-analysis as a research method.

Meta-analysis is a method of data analysis applied to summarizing research findings, both quantitative and qualitative summaries of individual studies. Rooted in the fundamental values of the scientific enterprise of replicability, causal and correlational analysis, it is useful for answering three general questions: What is the central tendency or typical study outcome? How much variability exists among study outcomes? What is the explanation of the variability? Advanced statistical and mathematical techniques are being used in counting studies with significant and non-significant findings; in combining effect size estimates based on fixed- or random-effects models; in the use of general linear models; and, Bayesian procedures. The methodology of meta-analysis and suggestions for publications are presented as well as appropriate software programs available for use in the meta-analytic process are explored. The benefits of meta-analysis as a research method in effecting health policy development is offered as a pragmatic perspective for future consideration. Conclusions have implications for the use of meta-analysis as a teaching strategy and as a methodology in nursing research and other applied sciences. In view of the rapid pace of knowledge development and increased public demand for accountability, meta-analysis offers an opportunity for organizing phenomena which gives direction for provision of quality health care.

Humans↗
PubMed1998

Change-point analysis of neuron spike train data.

In many medical experiments, data are collected across time, over a number of similar trials, or over a number of experimental units. As is the case of neuron spike train studies, these data may be in the form of counts of events per unit of time. These counts may be correlated within each trial. It is often of interest to know if the introduction of an intervention, such as the application of a stimulus, affects the distribution of the counts over the course of the experiment. In such investigations, each trial generates a sequence of data that may or may not contain a change in distribution at some point in time. Each sequence of integer counts can be viewed as arising from a Poisson process and are therefore independently distributed or as an integer-valued time series that allows for correlations between these counts. The main aim of this paper is to show how the ensemble of sample paths may be used to make inference about the distribution of the instantaneous times of change in a given population. This will be accomplished using a Bayesian hierarchical model for these change-points in time. A bonus of these models is they also allow for inference about the probability of a change in each unit and the magnitude of the effects, if any. The use of such change-point models on integer-valued time series is illustrated on neuron spike train data, although the methods can be applied to other situations where integer-valued processes arise.

Action Potentials↗
PubMed2025

Bayesian identification of differentially expressed isoforms using a novel joint model of RNA-seq data.

We develop a Bayesian approach, BayesIso, to identify differentially expressed isoforms from RNA-seq data. The approach features a novel joint model of the sample variability and the deferential state of isoforms. Specifically, the within-sample variability and the between-sample variability of each isoform are modeled by a Poisson-Lognormal model and a Gamma-Gamma model, respectively. Using a Bayesian framework, the differential state of each isoform and the model parameters are jointly estimated by a Markov Chain Monte Carlo (MCMC) method. Extensive studies using simulation and real data demonstrate that BayesIso can effectively detect isoforms of less differentially expressed and differential transcripts for genes with multiple isoforms. We applied the approach to breast cancer RNA-seq data and uncovered a unique set of isoforms that form key pathways associated with breast cancer recurrence. First, PI3K/AKT/mTOR signaling and PTEN signaling pathways are identified as being involved in breast cancer development. Further integrated with protein-protein interaction data, pathways of Jak-STAT, mTOR, MAPK and Wnt signaling are revealed in association with breast cancer recurrence. Finally, several pathways are activated in the early recurrence of breast cancer. In tumors that occur early, members of pathways of cellular metabolism and cell cycle (such as CD36 and TOP2A) are upregulated, while immune response genes such as NFATC1 are downregulated.

Humans↗
PubMed2025

Comparing artificial and convolutional neural networks with traditional models for Genomic prediction in wheat.

With the rapid development of sequencing technology, the application of genomic prediction has become more and more common in breeding schemes of livestocks and crops. Selecting an appropriate statistical model is of central importance to achieve high prediction accuracy. Recently, machine learning models have been expected to upgrade genomic prediction into a new era. However, the perspective still suffers from lack of evidence that machine learning models can generally outperform the traditional ones on empirical data sets. In this study, we compared two machine learning models based on artificial neural network (ANN) and convolutional neural network (CNN) with four traditional models, including genomic best linear unbiased prediction (GBLUP), Bayesian ridge regression (BRR), BayesA and BayesB, using three published data sets for grain yield in wheat. For each model, we considered two variants: modeling and ignoring the genotype-by-environment ([Formula: see text]) interaction. In the comparison, we considered two strategies of cross-validation: predicting genotypes that have not been evaluated in any environment (CV1) and predicting genotypes that have been tested in other environments (CV2). Our results showed that traditional Bayesian models (BayesA, BayesB, and BRR) outperformed GBLUP, ANN and CNN when considering [Formula: see text] interaction. The accuracies of ANN and CNN were higher than traditional models only in CV1 and when [Formula: see text] interaction was ignored. It was also found that the performance of the two machine learning models was significantly affected by the interaction between the CV strategy and the way of treating the [Formula: see text] interaction, while that of the four traditional models was only influenced by whether the [Formula: see text] interaction was considered or not. Thus, machine learning models can be a powerful complementary to the traditional ones and their superiority may depend on the prediction scenario. Among the two machine learning models, we observed that the accuracy of ANN was higher than CNN in most cases, indicating that it is still challenging to adapt complex machine learning models such as CNN to genomic prediction.

ANN↗
PubMed2001

A comparison of a Bayesian vs. a frequentist method for profiling hospital performance.

The objective of this study was to compare the classification of hospitals as outcomes outliers using a commonly implemented frequentist statistical approach vs. an implementation of Bayesian hierarchical statistical models, using 30-day hospital-level mortality rates for a cohort of acute myocardial infarction patients as a test case. For the frequentist approach, a logistic regression model was constructed to predict mortality. For each hospital, a risk-adj usted mortality rate was computed. Those hospitals whose 95% confidence interval, around the risk-adjusted mortality rate, excludes the mean mortality rate were classified as outliers. With the Bayesian hierarchical models, three factors could vary: the profile of the typical patient (low, medium or high risk), the extent to which the mortality rate for the typical patient departed from average, and the probability that the mortality rate was indeed different by the specified amount. The agreement between the two methods was compared for different patient profiles, threshold differences from the average and probabilities. Only marginal agreement was shown between the Bayesian and frequentist approaches. In only five of the 27 comparisons was the kappa statistic at least 0.40. The remaining 22 comparisons demonstrated only marginal agreement between the two methods. Within the Bayesian framework, hospital classification clearly depended on patient profile, threshold and probability of exceeding the threshold. These inconsistencies raise questions about the validity of current methods for classifying hospital performance, and suggest a need for urgent research into which methods are most meaningful to clinicians, managers and the general public.

Aged↗
PubMed2026

Internal Bayesian precision modulates the neural representation of social attention: Disentangling implicit and explicit components via model-informed multivariate EEG analysis.

Social attention integrates sensory cues with high-level cognitive expectations, yet the generative mechanisms through which implicit orienting and explicit belief-driven modulation interact remain poorly understood. This ambiguity complicates the distinction between specialized social modules and domain-general attentional processes. We combined a dynamic cueing task with hierarchical Bayesian modeling and model-informed multivariate EEG decoding to address this. Behavioral results revealed a computational double dissociation: symbolic arrow cues elicited heterogeneous strategies, whereas averted gaze recruited a consistent, surprise-driven computational phenotype. At the neural level, time-resolved decoding and temporal generalization revealed a critical representational shift starting approximately 400 ms post-cue. Initial activity related to physical cue features was rapidly replaced by stable neural templates of predicted spatial intent. Crucially, topographical activation patterns showed that this intentional template, characterized by a lateralized temporo-occipital distribution, emerged exclusively under high internal certainty. Furthermore, partial representational similarity analysis demonstrated that late-stage neural manifolds were overwhelmingly organized around integrated spatial goals rather than isolated sensory or motivational signals. These findings suggest that social attention is a specialized generative process, where internal certainty modulates the transformation of social perceptions into actionable top-down intentions.

Bayesian computational modeling↗
PubMed2001

Evolutionary HMMs: a Bayesian approach to multiple alignment.

MOTIVATION: We review proposed syntheses of probabilistic sequence alignment, profiling and phylogeny. We develop a multiple alignment algorithm for Bayesian inference in the links model proposed by Thorne et al. (1991, J. Mol. Evol., 33, 114-124). The algorithm, described in detail in Section 3, samples from and/or maximizes the posterior distribution over multiple alignments for any number of DNA or protein sequences, conditioned on a phylogenetic tree. The individual sampling and maximization steps of the algorithm require no more computational resources than pairwise alignment. METHODS: We present a software implementation (Handel) of our algorithm and report test results on (i) simulated data sets and (ii) the structurally informed protein alignments of BAliBASE (Thompson et al., 1999, Nucleic Acids Res., 27, 2682-2690). RESULTS: We find that the mean sum-of-pairs score (a measure of residue-pair correspondence) for the BAliBASE alignments is only 13% lower for Handelthan for CLUSTALW(Thompson et al., 1994, Nucleic Acids Res., 22, 4673-4680), despite the relative simplicity of the links model (CLUSTALW uses affine gap scores and increased penalties for indels in hydrophobic regions). With reference to these benchmarks, we discuss potential improvements to the links model and implications for Bayesian multiple alignment and phylogenetic profiling. AVAILABILITY: The source code to Handelis freely distributed on the Internet at http://www.biowiki.org/Handel under the terms of the GNU Public License (GPL, 2000, http://www.fsf.org./copyleft/gpl.html).

Algorithms↗
PubMed1998

Prediction of serum vancomycin concentrations using one-, two- and three-compartment models with implemented population pharmacokinetic parameters and with the Bayesian method.

Although previous studies have shown that vancomycin has a complicated pharmacokinetic profile requiring description using a two- or, better, three-compartment model, until recently predictions of serum vancomycin concentrations have been mainly based on one- or two-compartment models using computer software packages. In this study, we have predicted serum vancomycin concentrations in 59 patients using one-, two- and three-compartment models with implemented population pharmacokinetic parameters in the Abbott PKS program and by use of the Bayesian method. The percentage errors of predictions made using the one-compartment model were smaller when either the Bayesian method or implemented population pharmacokinetic parameters were used (medians of -8.61% and -9.49%, respectively). Predictions using the one-compartment model with the Bayesian method were less biased (median of -1.52 microgmL(-1). The best predictions were those made using the three-compartment model with the Bayesian method-they were most accurate (median of 3.40 microgmL(-1) and highly precise (median of 11.53 microg(2)mL(-1)). The results suggest that predictions made using the one-compartment model with implemented population pharmacokinetic parameters are preferable if no samples are available, otherwise predictions made using the three-compartment model with the Bayesian method are preferable. The results also supported our previous argument that the greater the number of compartments involved in individualization, the better the predictions obtained using the Bayesian method.

Adult↗
PubMed1997

Predictive variable selection for the multivariate linear model.

We develop a predictive Bayesian approach to variable selection in the multivariate linear model. A criterion derived from the Bayesian predictive density is proposed and a calibration is provided for it. Reference and informative priors are discussed, and an automated method that focuses on the response variable is proposed for specifying informative priors for the regression parameters. Relationships between the proposed criterion and other several well-known criteria are examined. Illustrative examples involving real data are given to demonstrate the methodology.

Asbestos↗
PubMed1996

Applications of a mixture survival model with covariates to the analysis of a depression prevention trial.

This paper presents a case study of model selection for survival analysis data. We use an approximate Bayesian method for model selection based on assessing the posterior probability of competing models given the data. We introduce the Schwarz criteria, an approximation to the logarithm of the Bayes factor, to provide an indication of evidence in favour of one model compared to another. Specifically, in the context of a depression prevention clinical trial we evaluate the efficacy of treatment in preventing or delaying the time to recurrence of depression, and evaluate how differences in the survival distributions between the two treatment groups depend on explanatory variables of interest. This investigation is based on a mixture survival model that explicitly incorporates the possibility of a surviving fraction.

Antidepressive Agents, Tricyclic↗
PubMed2001

Pharmacokinetic and pharmacodynamic modeling of anidulafungin (LY303366): reappraisal of its efficacy in neutropenic animal models of opportunistic mycoses using optimal plasma sampling.

The compartmental pharmacokinetics of anidulafungin (VER-002; formerly LY303366) in plasma were characterized with normal rabbits, and the relationships between drug concentrations and antifungal efficacy were assessed in clinically applicable infection models in persistently neutropenic animals. At intravenous dosages ranging from 0.1 to 20 mg/kg of body weight, anidulafungin demonstrated linear plasma pharmacokinetics that fitted best to a three-compartment open pharmacokinetic model. Following administration over 7 days, the mean (+/- standard error of the mean) peak plasma concentration (C(max)) increased from 0.46 +/- 0.02 microg/ml at 0.1 mg/kg to 63.02 +/- 2.93 microg/ml at 20 mg/kg, and the mean area under the concentration-time curve from 0 h to infinity (AUC(0-infinity)) rose from 0.71 +/- 0.04 to 208.80 +/- 24.21 microg. h/ml. The mean apparent volume of distribution at steady state (V(ss)) ranged from 0.953 +/- 0.05 to 1.636 +/- 0.22 liter/kg (nonsignificant [NS]), and clearance ranged from 0.107 +/- 0.01 to 0.149 +/- 0.00 liter/kg/h (NS). Except for a significant prolongation of the terminal half-life and a trend toward an increased V(ss) at the higher end of the dosage range after multiple doses, no significant differences in pharmacokinetic parameters were noted in comparison to single-dose administration. Concentrations in tissue at trough after multiple dosing (0.1 to 10 mg/kg/day) were highest in lung and liver (0.85 +/- 0.16 to 32.64 +/- 2.03 and 0.32 +/- 0.05 to 43.76 +/- 1.62 microg/g, respectively), followed by spleen and kidney (0.24 +/- 0.65 to 21.74 +/- 1.86 and <0.20 to 16.92 +/- 0.56, respectively). Measurable concentrations in brain tissue were found at dosages of > or =0.5 mg/kg (0.24 +/- 0.02 to 3.90 +/- 0.25). Implementation of optimal plasma sampling in persistently neutropenic rabbit infection models of disseminated candidiasis and pulmonary aspergillosis based on the Bayesian approach and model parameters from normal animals as priors revealed a significantly slower clearance (P < 0.05 for all dosage groups) with a trend toward higher AUC(0-24) values, higher plasma concentrations at the end of the dosing interval, and a smaller volume of distribution (P < 0.05 to 0.193 for the various comparisons among dosage groups). Pharmacodynamic modeling using the residual fungal tissue burden in the main target sites as the primary endpoint and C(max), AUC(0-24), time during the dosing interval of 24 h with plasma drug concentration equaling or exceeding the MIC or the minimum fungicidal concentration for the isolate, and tissue concentrations as pharmacodynamic parameters showed predictable pharmacokinetic-pharmacodynamic relationships in experimental disseminated candidiasis that fitted well with an inhibitory sigmoid maximum effect pharmacodynamic model (r(2), 0.492 to 0.819). However, no concentration-effect relationships were observed in experimental pulmonary aspergillosis using the residual fungal burden in lung tissue and survival as parameters of antifungal efficacy. Implementation of optimal plasma sampling in discriminative animal models of invasive fungal infections and pharmacodynamic modeling is a novel approach that holds promise of improving and accelerating our understanding of the action of antifungal compounds in vivo.

Anidulafungin↗
PubMed1993

Improved frequency resolution in multidimensional constant-time experiments by multidimensional Bayesian analysis.

The resolution of spectral frequencies in NMR data obtained from discrete Fourier transformation (DFT) along D constant-time dimensions can be improved significantly through extrapolation of the D-dimensional free induction decay (FID) by multidimensional Bayesian analysis. Starting from Bayesian probability theory for parameter estimation and model detection of one-dimensional time-domain data [Bretthorst, (1990) J. Magn. Reson., 88, 533-551; 552-570; 571-595], a theory for the D-dimensional case has been developed and implemented in an algorithm called BAMBAM (BAyesian Model Building Algorithm in Multidimensions). BAMBAM finds the most probable sinusoidal model to account for the systematic portion of any D-dimensional stationary FID. According to the parameters estimated by the algorithm, the FID is extrapolated in D dimensions prior to apodization and Fourier transformation. Multidimensional Bayesian analysis allows for the detection of signals not resolved by the DFT alone or even by sequential one-dimensional extrapolation from mirror-image linear prediction prior to the DFT. The procedure has been tested with a theoretical two-dimensional dataset and with four-dimensional HN(CO)CAHA (Kay et al. (1992) J. Magn. Reson., 98, 443-450) data from a small protein (8 kDa) where BAMBAM was applied to the 13C alpha and H alpha constant-time dimensions.

Algorithms↗
PubMed1996

A limited sampling method to estimate methotrexate pharmacokinetics in patients with rheumatoid arthritis using a Bayesian approach and the population data modeling program P-PHARM.

This paper describes a methodology to calculate methotrexate (MTX) pharmacokinetic parameters after intramuscular administration using two samples and the population parameters. Total and free MTX were measured over a 36-h period in 56 rheumatoid arthritis patients; 14 patients were studied after a two-dose scheme at 15-day intervals. The Hill equation was used to relate the free MTX to the total MTX changes in plasma concentrations, and a two-compartment open model was used to fit the total MTX plasma concentrations. A non-linear mixed effect procedure was used to estimate the population parameters and to explore the interindividual variability in relation to the following covariables: age, weight, height, haemoglobin, erythrocyte sedimentation rate, platelet count, creatinine clearance, rheumatoid factor, C-reactive protein, swelling joint count, and Ritchie's articular index. Population parameters were evaluated for 40 patients using a three-step approach. The population average parameters and the interindividual variabilities expressed as coefficients of variation (CV%) were: CL, 6.94 l center dot h-1 (20.5%); V, 34.8 l (32.2%); k12, 0.0838 h-1 (47.7%); k21, 0.0769 h-1 (61.6%); ka, 4.31 h-1 (58%); Emax, 1.12 mu mol center dot l-1 (19.7%); gamma, 0.932 (12.3%); and EC50, 2.14 mu mol center dot l-1 (27.3%). Thirty additional data sets (16 new patients and 14 patients of the previous population but treated on a separate occasion) were used to evaluate the predictive performance of the population parameters. Twelve blood samples were collected from each individual in order to calculate individual parameters using standard fitting procedures. These values were compared to the ones estimated using a Bayesian approach with population parameters as a priori information together with two samples, selected from the individual observations. The results show that the bias was not statistically different from zero and the precision of these parameters was excellent.

Adult↗