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 1,027 records · Page 57Linked to original sources
PubMed2004

Experimentally optimal nu in support vector regression for different noise models and parameter settings.

In Support Vector (SV) regression, a parameter nu controls the number of Support Vectors and the number of points that come to lie outside of the so-called epsilon-insensitive tube. For various noise models and SV parameter settings, we experimentally determine the values of nu that lead to the lowest generalization error. We find good agreement with the values that had previously been predicted by a theoretical argument based on the asymptotic efficiency of a simplified model of SV regression. As a side effect of the experiments, valuable information about the generalization behavior of the remaining SVM parameters and their dependencies is gained. The experimental findings are valid even for complex 'real-world' data sets. Based on our results on the role of the nu-SVM parameters, we discuss various model selection methods.

Generalization, Psychological↗
PubMed1996

Reduced variation at concertina, a heterochromatic locus in Drosophila.

In Drosophila melanogaster and closely related species, polymorphism has been shown to be reduced at loci located in regions of low recombination on the X chromosome and on the fourth chromosome, which does not normally recombine. This positive correlation between nucleotide polymorphism level and recombination rate is not predicted by standard neutral theory and therefore must result from natural selection and genetic hitchhiking along the chromosomes. We report here the near-complete absence of variation at concertina (cta), a locus located in the beta-heterochromatic base of chromosome 2L, a region of strongly reduced recombination. A 1.2 kilobase region containing coding regions and introns was sequenced from each of nine lines of D. melanogaster and nine lines of D. simulans representing worldwide collections. Variation is significantly reduced in eta in both species compared with other available loci on the same chromosome. Two analyses of background selection demonstrate that the reduction in variation at cta. considered in combination with other loci on chromosome 2L or alone, is consistent with the background selection model.

Animals↗
PubMed2006

A note on the asymptotic distribution of likelihood ratio tests to test variance components.

When using maximum likelihood methods to estimate genetic and environmental components of (co)variance, it is common to test hypotheses using likelihood ratio tests, since such tests have desirable asymptotic properties. In particular, the standard likelihood ratio test statistic is assumed asymptotically to follow a chi2 distribution with degrees of freedom equal to the number of parameters tested. Using the relationship between least squares and maximum likelihood estimators for balanced designs, it is shown why the asymptotic distribution of the likelihood ratio test for variance components does not follow a chi2 distribution with degrees of freedom equal to the number of parameters tested when the null hypothesis is true. Instead, the distribution of the likelihood ratio test is a mixture of chi2 distributions with different degrees of freedom. Implications for testing variance components in twin designs and for quantitative trait loci mapping are discussed. The appropriate distribution of the likelihood ratio test statistic should be used in hypothesis testing and model selection.

Chromosome Mapping↗
PubMed1983

Low-dose-rate extrapolation using the multistage model.

The distribution of the maximum likelihood estimates of virtually safe levels of exposure to environmental chemicals is derived by using large-sample theory and Monte Carlo simulation according to the Armitage-Doll multistage model. Using historical dose-response we develop a set of 33 two-stage models upon which we base our conclusions. The large-sample distributions of the virtually safe dose are normal for cases in which the multistage-model parameters have nonzero expectation, and are skewed in other cases. The large-sample theory does not provide a good approximation of the distribution observed for small bioassays when Monte Carlo simulation is used. The constrained nature of the multistage-model parameters leads to bimodal distributions for small bioassays. The two modes are the direct result of estimating the linear parameter in the multistage model; the lower mode results from estimating this parameter to be nonzero, and the upper mode from estimating it to be zero. The results of this research emphasize the need for incorporation of the biological theory in the model-selection process.

Biological Assay↗
PubMed2004

Linkage disequilibrium as a signature of selective sweeps.

The hitchhiking effect of a beneficial mutation, or a selective sweep, generates a unique distribution of allele frequencies and spatial distribution of polymorphic sites. A composite-likelihood test was previously designed to detect these signatures of a selective sweep, solely on the basis of the spatial distribution and marginal allele frequencies of polymorphisms. As an excess of linkage disequilibrium (LD) is also known to be a strong signature of a selective sweep, we investigate how much statistical power is increased by the inclusion of information regarding LD. The expected pattern of LD is predicted by a genealogical approach. Both theory and simulation suggest that strong LD is generated in narrow regions at both sides of the location of beneficial mutation. However, a lack of LD is expected across the two sides. We explore various ways to detect this signature of selective sweeps by statistical tests. A new composite-likelihood method is proposed to incorporate information regarding LD. This method enables us to detect selective sweeps and estimate the parameters of the selection model better than the previous composite-likelihood method that does not take LD into account. However, the improvement made by including LD is rather small, suggesting that most of the relevant information regarding selective sweeps is captured by the spatial distribution and marginal allele frequencies of polymorphisms.

Computer Simulation↗
PubMed1996

Learning temporal probabilistic causal models from longitudinal data.

Medical problems often require the analysis and interpretation of large collections of longitudinal data in terms of a structural model of the underlying physiological behavior. A suitable way to deal with this problem is to identify a temporal causal model that may effectively explain the patterns observed in the data. Here we will concentrate on probabilistic models, that provide a convenient framework to represent and manage underspecified information; in particular, we will consider the class of Causal Probabilistic Networks (CPN). We propose a method to perform structural learning of CPNs representing time-series through model selection. Starting from a set of plausible causal structures and a collection of possibly incomplete longitudinal data, we apply a learning algorithm to extract from the data the conditional probabilities describing each model. The models are then ranked according to their performance in reconstructing the original time-series, using several scoring functions, based on one-step ahead predictions. In this paper we describe the proposed methodology through an example taken from the diabetes monitoring domain. The selection process is applied to a set of input-output models that generalize the class of ARX models, where the inputs are the insulin and meal intakes and the outputs are the blood glucose levels. Although the physiological process underlying this particular application is characterized by strong non-linearities and low data reliability, we show that it is possible to obtain meaningful results, in terms of conditional probability learning and model ranking power.

Algorithms↗
PubMed2004

Host ecology determines the relative fitness of virus genotypes in mixed-genotype nucleopolyhedrovirus infections.

Mixed-genotype infections are common in many natural host-parasite interactions. Classical kin-selection models predict that single-genotype infections can exploit host resources prudently to maximize fitness, but that selection favours rapid exploitation when co-infecting genotypes share limited host resources. However, theory has outpaced evidence: we require empirical studies of pathogen genotypes that naturally co-infect hosts. Do genotypes actually compete within hosts? Can host ecology affect the outcome of co-infection? We posed both questions by comparing traits of infections in which two baculovirus genotypes were fed to hosts alongside inocula of the same or a different genotype. The host, Panolis flammea, is a herbivore of Pinus sylvestris and Pi. contorta. The pathogen, PfNPV (a nucleopolyhedrovirus), occurs naturally as mixtures of genotypes that differ, when isolated, in pathogenicity, speed of kill and yield. Single-genotype infection traits failed to predict the 'winning' genotypes in co-infections. Co-infections infected and caused lethal disease in more hosts, and produced high yields, relative to single-genotype infections. The need to share with nonkin did not cause fitness costs to either genotype. In fact, in hosts feeding on Pi. sylvestris, one genotype gained increased yields in mixed-genotype infections. These results are discussed in relation to theory surrounding adaptive responses to competition with nonkin for limited resources.

Animals↗
PubMed1999

Applying Bayesian ideas to the development of medical guidelines.

Measurements of the quality of health care, in particular the underuse and overuse of medical therapies and diagnostic tests, often involve employment of medical practice guidelines to assess the appropriateness of treatments. This paper presents a case study of a Bayesian analysis for the development of medical guidelines based on expert opinion, using ordinal categorical rater data. We develop guidelines for the use of coronary angiography following an acute myocardial infarction (AMI) for 890 clinical indications using statistical models fit to appropriateness ratings obtained from a nine-member expert panel. The main foci of our analyses were on the estimation of an appropriateness score for each of the clinical indications, an associated measure of precision, and functions of the underlying score. We considered two classes of models that assume the ratings are either in the form of grouped normal data or are ungrouped variables arising from a normal distribution, while permitting rater effects and indication heterogeneity in both. We estimated models using Markov chain Monte Carlo methods and constructed indices quantifying appropriateness based on posterior probabilities of selected model parameters. We compared our model-based approach to the standard approach currently employed in medical guideline development and found that the standard approach correctly identified 99 per cent of the appropriate indications while overestimating appropriateness 18 per cent of the time compared to our model-based approach.

Bayes Theorem↗
PubMed1993

The importance of modeling interoccasion variability in population pharmacokinetic analyses.

Individual pharmacokinetic parameters may change randomly between study occasions. Analysis of simulated data with NONMEM shows that ignoring such interoccasion variability (IOV) may result in biased population parameter estimates. Particular parameters affected and the extent to which they are biased depend on study design and the magnitude of IOV and interindividual variability. Neglecting IOV also results in a high incidence of statistically significant spurious period effects. Perhaps most important, ignoring IOV can lead to a falsely optimistic impression of the potential value of therapeutic drug monitoring. A model incorporating IOV was developed and its performance in the presence and absence of IOV was evaluated. The IOV model performs well with respect to both model selection and population parameter estimation in all circumstances studied. Analysis of two real data examples using this model reveals significant IOV in all parameters for both drugs and supports the simulation findings for the case that IOV is ignored: predictable biases occur in parameter estimates and previously nonexistent period effects are found.

Computer Simulation↗
PubMed1992

Marital status, fatness and obesity.

Marital status is related to morbidity and mortality, with married people healthier and at lower risk of death than those who are unmarried (especially among men). However, the relationship between marital status and obesity is not well established. Role theory suggests through a marital causation model that people in the marital role are more likely to be obese, and through a marital selection model that people in the marital role are less likely to be obese because of stigmatization. The martial causation model was examined using data from the National Survey of Personal Health Practices and Consequences, a cross-sectional national telephone survey of 3025 adults age 20-64 in the United States. Sequential regression analyses revealed that married men were significantly fatter and more likely to be obese than never married or previously married men, even when demographic, social, and physical variables were controlled. By contrast, marital status was not significantly associated with fatness or obesity among women when other variables were controlled. The marital role appears to influence fatness and obesity among men, but not women.

Cross-Sectional Studies↗
PubMed2003

Estimating group mortality and paraplegia rates after thoracoabdominal aortic aneurysm repair.

BACKGROUND: Most clinical studies regarding thoracoabdominal aortic aneurysm (TAAA) surgery are retrospective comparisons involving heterogeneous groups of patients. Risk models that evaluate susceptibility bias enhance interpretation of these intergroup comparisons. The purpose of this analysis was to derive group risk models for mortality and paraplegia after TAAA repair. METHODS: Data regarding 1,220 consecutive patients undergoing TAAA repair were analyzed via multiple logistic regression with stepwise model selection. Categorical preoperative risk factors that predicted 30-day mortality and paraplegia were used to develop risk models. RESULTS: Fifty-eight patients (4.8%) died within 30 days and 56 patients (4.6%) developed paraplegia or paraparesis. Predictors of mortality were rupture, renal insufficiency, symptomatic aneurysms, and Crawford extent II repairs. Extent of repair and acute presentation were predictors of paraplegia. The derived risk models estimated mortality and paraplegia rates that correlated well with actual frequencies reported in other contemporary series (regression slopes = 0.87 and 1.06, respectively). CONCLUSIONS: The derived risk models accurately estimate paraplegia and mortality rates in groups of patients. Prospective model validation will be required to confirm their accuracy.

Adolescent↗
PubMed1998

Angiotensin-converting enzyme gene polymorphism is associated with coronary heart disease in non-insulin-dependent diabetic patients evaluated for 9 years.

The insertion/deletion (I/D) polymorphism of the human angiotensin-converting enzyme (ACE) gene is a major determinant of circulating ACE levels. Recent studies have found the ACE D allele to be associated with an increased risk for coronary heart disease (CHD) in diabetic and nondiabetic subjects. This association has not been evaluated in prospective studies. We therefore studied the relationship between ACE gene I/D polymorphism and CHD in patients with non-insulin-dependent diabetes mellitus (NIDDM) evaluated for 9 years. The I/D polymorphism was determined by polymerase chain reaction (PCR). Overestimation of the frequency of the DD genotype was eliminated by insertion-specific primers and inclusion of 5% dimethylsulfoxide (DMSO). Eighty-three patients were evaluated for a mean period of 9.1 years (range, 7.4 to 10.5). Among them, 64 patients showed no CHD at entry. During the follow-up period, 21 patients (37.5%) developed CHD. The systolic blood pressure (P = .046), fasting blood glucose (P < .01), and prevalence of hypertension (P < .001) increased, while high-density lipoprotein (HDL) cholesterol (P < .001) decreased. Patients who developed CHD were older than those who did not; the mean age was 59.3 and 53.2 years, respectively (P = .003). The prevalence of albuminuria at follow-up examination was higher in CHD subjects versus non-CHD subjects (61.9% v 20.9%, P = .012). The D allele of the ACE gene was significantly more frequent in subjects with CHD versus those without CHD in both follow-up (P = .028, chi2 test) and cross-sectional (P = .033, chi2 test) settings. No difference could be detected between the three genotypes in age, body mass index (BMI), blood pressure, or plasma lipid levels. In our logistic regression analysis, the best model selected the DD genotype (P = .0105) and age (P = .0407) as significant risk factors for CHD. This model classified 89% of the subjects correctly. In conclusion, this 9-year prospective study supports the hypothesis that the ACE I/D polymorphism is an important and independent risk factor for CHD in patients with NIDDM.

Adult↗
PubMed2000

General linear models for multicenter clinical trials.

The use of generalized linear models (GLM) in the analysis of multicenter clinical trials is commonplace. Currently, diverse opinion exists as to the proper modeling approach in this setting. This paper provides a description of the technical backdrop underlying the model selection and discusses the pros and cons of the usual parameter estimators. Arguments are given favoring a model incorporating a term for interaction effect while attenuating the effects of small centers.

Clinical Trials as Topic↗
PubMed2001

[Ethical decision models in nursing and medicine].

A literature review on ethical decision making models shows that there is a great number of monodisciplinary ethical models in nursing and in medicine and new ones are constantly being developed. It is, however, to be questioned if by using these models divergent possibilities and consequences arise for practice. To answer this question this article deals with two selected models in ethical decision making. The model by Illhardt (1998) derives from a physician's viewpoint and understanding. The procedure from Tschudin (1996) was developed on the background of nursing. The practical application of these two models in this case study shows common ground as well as differences and interrelated effects on the decision making process. In relation to this case study the development of specific ethical decision making models for nursing is therefore questionable, most of all if an interdisciplinary implementation is the actual aim.

Aged↗
PubMed1984

Statistical properties of single sodium channels.

Single channel currents were obtained from voltage-activated sodium channels in outside-out patches of tissue-cultured GH3 cells, a clonal line from rat pituitary gland. In membrane patches where the probability of overlapping openings was low, the open time histograms were well fit by a single exponential. Most analysis was done on a patch with exactly one channel. We found no evidence for multiple open states at -25 and -40 mV, since open times, burst durations, and autocorrelation functions were time independent. Amplitude histograms showed no evidence of multiple conductance levels. We fit the gating with 25 different time-homogeneous Markov chain models having up to five states, using a maximum likelihood procedure to estimate the rate constants. For selected models, this procedure yielded excellent predictions for open time, closed time, and first latency density functions, as well as the probability of the channel being open after a step depolarization, the burst duration distribution, autocorrelation, and the distribution of number of openings per record. The models were compared statistically using likelihood ratio tests and Akaike's information criterion. Acceptable models allowed inactivation from closed states, as well as from the open state. Among the models eliminated as unacceptable by this survey were the Hodgkin-Huxley model and any model requiring a channel to open before inactivating.

Animals↗
PubMed2005

In vivo selection of engineered homing endonucleases using double-strand break induced homologous recombination.

Homing endonucleases, endonucleases capable of recognizing long DNA sequences, have been shown to be a tool of choice for precise and efficient genome engineering. Consequently, the possibility to engineer novel endonucleases with tailored specificities is under strong investigation. In this report, we present a simple and efficient method to select meganucleases from libraries of variants, based on their cleavage properties. The method has the advantage of directly selecting for the ability to induce double-strand break induced homologous recombination in a eukaryotic environment. Model selections demonstrated high levels of enrichments. Moreover, this method compared favorably with phage display for enrichment of active mutants from a mutant library. This approach makes possible the exploration of large sequence spaces and thereby represents a valuable tool for genome engineering.

Binding Sites↗
PubMed2003

Does recombination shape the distribution and evolution of tandemly arrayed genes (TAGs) in the Arabidopsis thaliana genome?

Tandemly arrayed genes (TAGs) are an important genomic component. However, most previous studies have focused on individual TAG families, and a broader characterization of their genomic distribution is not yet available. In this study, we examined the distribution of TAGs in the Arabidopsis thaliana genome and examined TAG density with relation to recombination rates. Recombination rates along A. thaliana chromosomes were estimated by comparing a genetic map with the genome sequence. Average recombination rates in A. thaliana are high, and rates vary more than threefold among chromosomal regions. Comparisons between TAG density and recombination indicate a positive correlation on chromosomes 1, 2, and 3. Moreover, there is a consistent centromeric effect. Relative to single-copy genes, TAGs are proportionally less frequent in centromeres than on chromosomal arms. We also examined several factors that have been proposed to affect the sequence evolution of TAG members. Sequence divergence is related to the number of members in the TAG, but genomic location has no obvious effect on TAG sequence divergence, nor does the presence of unrelated genes within a TAG. Overall, the distribution of TAGs in the genome is not consistent with theoretical models predicting the accumulation of repeats in regions of low recombination but may be consistent with stabilizing selection models of TAG evolution.

Arabidopsis↗