Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “model selection”

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 577 records · Page 32Linked to original sources
PubMed1972

Experimentally determined, two locus fitnesses of Drosophila melanogaster males.

A pilot demonstration of experimental fitness estimation utilizing pedigree data from D. melanogaster was described. Fitnesses were obtained for the male segregants of the two "complex loci," spineless-spineless aristapedia and radius incompletus-inturned. A good fit between selection model and data was obtained, perhaps because the experimental design largely eliminated concerns of frequency or density-dependent selection and assortative mating. The selection model assumed that net fitness was divided into two components: an "early" component operative from conception to the time of observation in the generation interval, and a late component operative from the time of observation to the time of gamete pool formation. These fitness components were roughly descriptive of the physiological components viability and reproductive ability. No simple systematic relationship between the fitness of two-locus genotypes and the marginal one-locus fitness emerged. These conclusions were based on relatively precise (by current standards) fitness estimates. It was argued that pedigree data is highly desirable for fitness estimation.

Age Factors↗
PubMed2006

Strength of evidence for density dependence in abundance time series of 1198 species.

Population limitation is a fundamental tenet of ecology, but the relative roles of exogenous and endogenous mechanisms remain unquantified for most species. Here we used multi-model inference (MMI), a form of model averaging, based on information theory (Akaike's Information Criterion) to evaluate the relative strength of evidence for density-dependent and density-independent population dynamical models in long-term abundance time series of 1198 species. We also compared the MMI results to more classic methods for detecting density dependence: Neyman-Pearson hypothesis-testing and best-model selection using the Bayesian Information Criterion or cross-validation. Using MMI on our large database, we show that density dependence is a pervasive feature of population dynamics (median MMI support for density dependence = 74.7-92.2%), and that this holds across widely different taxa. The weight of evidence for density dependence varied among species but increased consistently, with the number of generations monitored. Best-model selection methods yielded similar results to MMI (a density-dependent model was favored in 66.2-93.9% of species time series), while the hypothesis-testing methods detected density dependence less frequently (32.6-49.8%). There were no obvious differences in the prevalence of density dependence across major taxonomic groups under any of the statistical methods used. These results underscore the value of using multiple modes of analysis to quantify the relative empirical support for a set of working hypotheses that encompass a range of realistic population dynamical behaviors.

Ecosystem↗
PubMed2006

A graphical approach to relatedness inference.

The estimation of relatedness structure in natural populations using molecular marker data has become an important tool in population biology, resulting in a variety of estimation procedures for specific sampling scenarios. In this article a general approach is proposed, in which the detailed relationship structure, typically a pedigree graph or partition, is considered to be the object of inference. This makes available tools used in complex model selection theory which have demonstrated effectiveness. An important advantage of this approach is that it permits a fully Bayesian approach to the problem, providing a principled and accessible way to measure statistical error. The approach is demonstrated by applying the minimum description length principle. This technique is used in model selection to provide a rational way of comparing models of varying complexity. We show how the resulting score may be interpreted and applied as a Bayesian posterior density.

Bayes Theorem↗
PubMed2001

SCA7 mouse models show selective stabilization of mutant ataxin-7 and similar cellular responses in different neuronal cell types.

Accumulation of expanded polyglutamine proteins and selective pattern of neuronal loss are hallmarks of at least eight neurodegenerative disorders, including spinocerebellar ataxia type 7 (SCA7). We previously described SCA7 mice displaying neurodegeneration with progressive ataxin-7 accumulation in two cell types affected in the human pathology. We describe here a new transgenic model with a more widespread expression of mutant ataxin-7, including neuronal cell types unaffected in SCA7. In these mice a similar handling of mutant ataxin-7, including a cytoplasm to nucleus translocation and accumulation of N-terminal fragments, was observed in all neuronal populations studied. An extensive screen for chaperones, proteasomal subunits and transcription factors sequestered in nuclear inclusions (NIs) disclosed no pattern unique to neurons undergoing degeneration in SCA7. In particular, we found that the mouse TAF(II)30 subunit of the TFIID initiation complex is markedly accumulated in NIs, even though this protein does not contain a polyglutamine stretch. A striking discrepancy between mRNA and ataxin-7 levels in transgenic mice expressing the wild-type protein but not in those expressing the mutant one, indicates a selective stabilization of mutant ataxin-7, both in this model and the P7E/N model described previously. These mice therefore provide in vivo evidence that the polyglutamine expansion mutation can stabilize its target protein.

Animals↗
PubMed2006

Alternatives to statistical hypothesis testing in ecology: a guide to self teaching.

Statistical methods emphasizing formal hypothesis testing have dominated the analyses used by ecologists to gain insight from data. Here, we review alternatives to hypothesis testing including techniques for parameter estimation and model selection using likelihood and Bayesian techniques. These methods emphasize evaluation of weight of evidence for multiple hypotheses, multimodel inference, and use of prior information in analysis. We provide a tutorial for maximum likelihood estimation of model parameters and model selection using information theoretics, including a brief treatment of procedures for model comparison, model averaging, and use of data from multiple sources. We discuss the advantages of likelihood estimation, Bayesian analysis, and meta-analysis as ways to accumulate understanding across multiple studies. These statistical methods hold promise for new insight in ecology by encouraging thoughtful model building as part of inquiry, providing a unified framework for the empirical analysis of theoretical models, and by facilitating the formal accumulation of evidence bearing on fundamental questions.

Algorithms↗
PubMed2006

Simultaneous fine mapping of multiple closely linked quantitative trait Loci using combined linkage disequilibrium and linkage with a general pedigree.

Within a small region (e.g., <10 cM), there can be multiple quantitative trait loci (QTL) underlying phenotypes of a trait. Simultaneous fine mapping of closely linked QTL needs an efficient tool to remove confounded shade effects among QTL within such a small region. We propose a variance component method using combined linkage disequilibrium (LD) and linkage information and a reversible jump Markov chain Monte Carlo (MCMC) sampling for model selection. QTL identity-by-descent (IBD) coefficients between individuals are estimated by a hybrid MCMC combining the random walk and the meiosis Gibbs sampler. These coefficients are used in a mixed linear model and an empirical Bayesian procedure combines residual maximum likelihood (REML) to estimate QTL effects and a reversible jump MCMC that samples the number of QTL and the posterior QTL intensities across the tested region. Note that two MCMC processes are used, i.e., an (internal) MCMC for IBD estimation and an (external) MCMC for model selection. In a simulation study, the use of the multiple-QTL model clearly removes the shade effects between three closely linked QTL located at 1.125, 3.875, and 7.875 cM across the region of 10 cM, using 40 markers at 0.25-cM intervals. It is shown that the use of combined LD and linkage information gives much more useful information compared to using linkage information alone for both single- and multiple-QTL analyses. When using a lower marker density (11 markers at 1-cM intervals), the signal of the second QTL can disappear. Extreme values of past effective size (resulting in extreme levels of LD) decrease the mapping accuracy.

Chromosome Mapping↗
PubMed2005

Mathematical models assuming selective recruitment fitted to data for driver mortality and seat belt use in Japan.

Previous research has indicated that unbelted drivers are at higher risk of involvement in fatal crashes than belted drivers, suggesting selective recruitment that high-risk drivers are unlikely to become belt users. However, how the risk of involvement in fatal crashes among unbelted drivers varies according to the level of seat belt use among general drivers has yet to be clearly quantified. We, therefore, developed mathematical models describing the risk of fatal crashes in relation to seat belt use among the general public, and explored how these models fitted to changes in driver mortality and changes in observed seat belt use using Japanese data. Mortality data between 1979 and 1994 were obtained from vital statistics, and mortality data in the daytime and nighttime between 1980 and 2001 and belt use data between 1979 and 2001 were obtained from the National Police Agency. Regardless of the data set analyzed, exponential models, assuming that high-risk drivers would gradually become belt users in order of increasing risk as seat belt use among general motorists reached high levels, showed the best fit. Our models provide an insight into behavioral changes among high-risk drivers and support the selective recruitment hypothesis.

Accidents, Traffic↗
PubMed1996

Genetic versus phenotypic models of selection: can genetics be neglected in a long-term perspective?

Game theoretical concepts in evolutionary biology have been criticized by populations geneticists, because they neglect such crucial aspects as the mating system or the mode of inheritance. In fact, the dynamics of natural selection does not necessarily lead to a fitness maximum or an ESS if genetic constraints are taken into account. Yet, it may be premature to conclude that game theoretical concepts do not have a dynamical justification. The new paradigm of long-term evolution postulates that genetic constraints, which may be dominant in a short-term perspective, will in the long run disappear in the face of the ongoing influx of mutations. Two basic results (see Hammerstein; this issue) seem to reconcile the dynamical approach of long-term population genetics with the static approach of evolutionary game theory: (1) only populations at local fitness optima (Nash strategies) can be long-term stable; and (2) in monomorphic populations, evolutionary stability is necessary and sufficient to ensure long-term dynamic stability. The present paper has a double purpose. On the one hand, it is demonstrated by fairly general arguments that the scope of the results mentioned above extends to non-linear frequency dependent selection, to multiple loci, and to quite general mating systems. On the other hand, some limitations of the theory of long-term evolution will also be stressed: (1) there is little hope for a game theoretical characterization of stability in polymorphic populations; (2) many interesting systems do not admit long-term stable equilibria; and (3) even if a long-term stable equilibrium exists, it is not at all clear whether and how it is attainable by a series of gene substitution events.

Biological Evolution↗
PubMed1996

A technique for measuring epidemiologically useful features of birthweight distributions.

Birthweight distributions have been conceptualized as a predominant Gaussian distribution contaminated in the tails by an unspecified 'residual' distribution. Acknowledging this idea, we propose a technique for measuring certain features of birthweight distributions useful to epidemiologists: the mean and variance of the predominant distribution; the proportions of births in the low- and high-birthweight residual distributions, and the boundaries of support for these residual distributions. Our technique, based on an underlying multinomial sampling distribution, involves estimating parameters in a mixture model for the multinomial bin probabilities after having chosen the support of the residual distribution with a model selection criterion. A modest simulation study and experience with a few actual datasets indicate that use of a Bayesian information criterion (BIC) as model selection criterion is superior to use of Akaike's information criterion (AIC) in this application.

Analysis of Variance↗
PubMed2004

Optimizing dose selection with modeling and simulation: application to the vasopeptidase inhibitor M100240.

Dual inhibition of neutral endopeptidase 24.11 (NEP) and angiotensin-converting enzyme (ACE) has gained increasing interest in the treatment of hypertension, heart failure, and renoprotection. Specifically, M100240, the thioester of the dual ACE/NEP inhibitor MDL100,173, has been evaluated in the management of hypertension. A model-based analysis, including simulations, was employed to characterize the relationship between individual M100240 drug exposure and neurohormonal response and to optimize the dose selection for future clinical studies. Sixty-two healthy subjects and 189 hypertensive patients were studied after oral once-daily administration of 2.5, 5, 10, 25, or 50 mg M100240. Pharmacokinetic-biomarker and blood pressure response models were fitted to the data with the computer program NONMEM. A direct inhibitory E(max) model adequately described the relationship between MDL100,173 concentration and ACE activity. No clear concentration or dose-dependent NEP or blood pressure responses were evident. Given a target 90% ACE inhibition, simulation reveals that (1). 50 mg M100240 once daily produces adequate ACE inhibition 24 hours postdose in only 20% of subjects, and (2). higher and/or more frequent doses on the order of 25 mg three times daily or 50 mg twice daily are required to achieve the target ACE inhibition in at least 50% of patients over 24 hours. Insufficient blood pressure-lowering effects were observed in healthy subjects and hypertensive patients due to inadequate ACE and NEP inhibition with once-daily oral doses of up to 50 mg of M100240. Divided doses might provide target ACE inhibition in more patients.

Angiotensin-Converting Enzyme Inhibitors↗
PubMed2005

Evaluation of a selective neurectomy model for low urethral pressure incontinence in female dogs.

OBJECTIVE: To develop a model of low urethral pressure incontinence and compare the relative contributions of the pudendal and hypogastric nerves with urethral function by performing selective neurectomy and ovariohysterectomy in dogs. ANIMALS: 19 healthy Foxhounds. PROCEDURE: Dogs were allocated into 2 groups. The first group (10 dogs) underwent bilateral hypogastric neurectomy and ovariohysterectomy and subsequent bilateral pudendal neurectomy. The second group (9 dogs) underwent bilateral pudendal neurectomy and subsequent hypogastric neurectomy and ovariohysterectomy. Urethral pressure profilometry and leak point pressure (LPP) tests were performed before and after each neurectomy. RESULTS: Before surgery, mean +/- SD LPP and maximal urethral closure pressure (MUCP) in all dogs were 169.3 +/- 24.9 cm H2O and 108.3 +/- 19.3 cm H2O, respectively; these values decreased to 92.3 +/- 27 cm H2O and 60.7 +/- 20.0 cm H2O, respectively, after both selective neurectomy surgeries. There was a progressive decline of LPP after each neurectomy; however, MUCP decreased only after pudendal neurectomy. Fifteen dogs had mild clinical signs of urinary incontinence. All dogs appeared to have normal bladder function as indicated by posturing to void and consciously voiding a full stream of urine. Urinary tract infection did not develop in any dog. CONCLUSIONS AND CLINICAL RELEVANCE: Hypogastric and pudendal neurectomy and ovariohysterectomy caused a maximum decrease in LPP whereas pudendal neurectomy caused a maximum decrease in MUCP. IMPACT ON HUMAN MEDICINE: This model may be useful for evaluation of treatments for improving urinary control in postmenopausal women.

Animals↗
PubMed1997

The efficacy of different moisturizers on barrier recovery in hairless mice evaluated by non-invasive bioengineering methods. A model to select the potentially most effective product.

Moisturizers (emollients) are used frequently on normal and diseased skin. However, only few studies have examined their effects in dynamic situations and in more clinically relevant settings. We evaluated the effect of 4 commonly used products in a hairless mice model after acute skin barrier perturbation with acetone. The efficacy was evaluated by measurement of the transepidermal water loss (TEWL) and electrical conductance at various time intervals during barrier repair. The test products were compared with acetone-treated air-exposed controls allowed to recover otherwise normally and with a known irritant product, chlorhexidine cream 1%. Locobase was the most effective product in correcting barrier function and significantly improved barrier function during early stages of barrier recovery (< 6 h) without interfering with late stages of barrier recovery (> 6 h). The irritant control product, chlorhexidine cream 1%, delayed barrier recovery in the late stages. The model makes it possible to evaluate the combined effects of exogenous and endogenous components on barrier repair and to select the potentially most effective products before performing more cumbersome and time-consuming field studies.

Acetone↗
PubMed2006

Singularities affect dynamics of learning in neuromanifolds.

The parameter spaces of hierarchical systems such as multilayer perceptrons include singularities due to the symmetry and degeneration of hidden units. A parameter space forms a geometrical manifold, called the neuromanifold in the case of neural networks. Such a model is identified with a statistical model, and a Riemannian metric is given by the Fisher information matrix. However, the matrix degenerates at singularities. Such a singular structure is ubiquitous not only in multilayer perceptrons but also in the gaussian mixture probability densities, ARMA time-series model, and many other cases. The standard statistical paradigm of the Cramér-Rao theorem does not hold, and the singularity gives rise to strange behaviors in parameter estimation, hypothesis testing, Bayesian inference, model selection, and in particular, the dynamics of learning from examples. Prevailing theories so far have not paid much attention to the problem caused by singularity, relying only on ordinary statistical theories developed for regular (nonsingular) models. Only recently have researchers remarked on the effects of singularity, and theories are now being developed. This article gives an overview of the phenomena caused by the singularities of statistical manifolds related to multilayer perceptrons and gaussian mixtures. We demonstrate our recent results on these problems. Simple toy models are also used to show explicit solutions. We explain that the maximum likelihood estimator is no longer subject to the gaussian distribution even asymptotically, because the Fisher information matrix degenerates, that the model selection criteria such as AIC, BIC, and MDL fail to hold in these models, that a smooth Bayesian prior becomes singular in such models, and that the trajectories of dynamics of learning are strongly affected by the singularity, causing plateaus or slow manifolds in the parameter space. The natural gradient method is shown to perform well because it takes the singular geometrical structure into account. The generalization error and the training error are studied in some examples.

Journal Article↗
PubMed2000

Hierarchical bayesian models for regularisation in sequential learning

We show that a hierarchical Bayesian modeling approach allows us to perform regularization in sequential learning. We identify three inference levels within this hierarchy: model selection, parameter estimation, and noise estimation. In environments where data arrive sequentially, techniques such as cross validation to achieve regularization or model selection are not possible. The Bayesian approach, with extended Kalman filtering at the parameter estimation level, allows for regularization within a minimum variance framework. A multilayer perceptron is used to generate the extended Kalman filter nonlinear measurements mapping. We describe several algorithms at the noise estimation level that allow us to implement on-line regularization. We also show the theoretical links between adaptive noise estimation in extended Kalman filtering, multiple adaptive learning rates, and multiple smoothing regularization coefficients.

Journal Article↗