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 433 records · Page 24Linked to original sources
PubMed2005

Bayesian analysis for generalized linear models with nonignorably missing covariates.

We propose Bayesian methods for estimating parameters in generalized linear models (GLMs) with nonignorably missing covariate data. We show that when improper uniform priors are used for the regression coefficients, phi, of the multinomial selection model for the missing data mechanism, the resulting joint posterior will always be improper if (i) all missing covariates are discrete and an intercept is included in the selection model for the missing data mechanism, or (ii) at least one of the covariates is continuous and unbounded. This impropriety will result regardless of whether proper or improper priors are specified for the regression parameters, beta, of the GLM or the parameters, alpha, of the covariate distribution. To overcome this problem, we propose a novel class of proper priors for the regression coefficients, phi, in the selection model for the missing data mechanism. These priors are robust and computationally attractive in the sense that inferences about beta are not sensitive to the choice of the hyperparameters of the prior for phi and they facilitate a Gibbs sampling scheme that leads to accelerated convergence. In addition, we extend the model assessment criterion of Chen, Dey, and Ibrahim (2004a, Biometrika 91, 45-63), called the weighted L measure, to GLMs and missing data problems as well as extend the deviance information criterion (DIC) of Spiegelhalter et al. (2002, Journal of the Royal Statistical Society B 64, 583-639) for assessing whether the missing data mechanism is ignorable or nonignorable. A novel Markov chain Monte Carlo sampling algorithm is also developed for carrying out posterior computation. Several simulations are given to investigate the performance of the proposed Bayesian criteria as well as the sensitivity of the prior specification. Real datasets from a melanoma cancer clinical trial and a liver cancer study are presented to further illustrate the proposed methods.

Bayes Theorem↗
PubMed1991

Influence of a decision analysis model on selection of drug therapy.

The impact of a decision analysis model on pharmacists' preferences for a drug therapy was studied. Three hundred forty members of the American College of Clinical Pharmacy were randomly assigned to receive either a copy of a drug therapy review alone (control group) or the review plus a decision analysis of the same clinical problem (experimental group). The disease selected was pseudomembranous colitis, and the drugs to be considered were vancomycin, metronidazole, and bacitracin. The review and the decision analysis recommended metronidazole as the best agent. Each subject completed a short questionnaire, read the review or the review plus the analysis, and completed a follow-up questionnaire. Both questionnaires asked the subjects to rank the agents in order of preference. There were 164 usable responses, 86 from the control group and 78 from the experimental group; the total response rate was 48.2%. The difference in the proportion of respondents in each group who chose metronidazole as the most preferred agent, both before and after the intervention, was not significant. Of the 40 experimental-group subjects who ranked metronidazole as their second or third choice in the pretest, 16 (40%) ranked it as the most preferred agent in the posttest. Ten of these 16 stated that the model influenced their decision. Decision analysis plus a drug therapy review had no greater impact on pharmacists' opinions on the selection of drug therapy for pseudomembranous colitis than did the review alone. The model did influence some of those who changed their opinion.

Attitude of Health Personnel↗
PubMed1986

Spin glass model of learning by selection.

A model of learning by selection is described at the level of neuronal networks. It is formally related to statistical mechanics with the aim to describe memory storage during development and in the adult. Networks with symmetric interactions have been shown to function as content-addressable memories, but the present approach differs from previous instructive models. Four biologically relevant aspects are treated--initial state before learning, synaptic sign changes, hierarchical categorization of stored patterns, and synaptic learning rule. Several of the hypotheses are tested numerically. Starting from the limit case of random connections (spin glass), selection is viewed as pruning of a complex tree of states generated with maximal parsimony of genetic information.

Learning↗
PubMed2002

A prediction model for selecting patients undergoing in vitro fertilization for elective single embryo transfer.

OBJECTIVE: Construction of a prediction model to enable the selection of patients for elective single ET. DESIGN: Retrospective cohort study. SETTING: Fertility center in a tertiary referral university hospital. PATIENT(S): Six hundred forty-two women undergoing their first IVF treatment cycle in which no more than two embryos were transferred. INTERVENTION(S): Database analysis. MAIN OUTCOME MEASURE(S): Ongoing pregnancy and multiple pregnancy. RESULT(S): In multivariate analysis, the best predictors for ongoing pregnancy were female age, the number of retrieved oocytes, the developmental stage score and the morphology score of the two best embryos available for transfer, and the day of transfer. Younger age and high quality of transferred embryos were the best predictors for increased risk of multiple pregnancy. The resulting model enables the calculation of probabilities of pregnancy and twin pregnancy. Depending on embryo quality, there is a threshold age under which the chance of singleton pregnancy is higher if one embryo is transferred compared with two embryos. CONCLUSION(S): Application of this model may enable a reduction in the chance of twin pregnancy without compromising singleton pregnancy rates in a subgroup of patients undergoing IVF.

Adult↗
PubMed2004

Engineering charge selectivity in model ion channels.

Most ion channel proteins exhibit some degree of charge selectivity, that is, an ability to conduct ions of one charge more efficiently than ions of the opposite charge. The structural origins of charge selectivity remain incompletely understood despite recent advances in the determination of cation-selective and anion-selective channel protein structures. Helix bundle channels formed via self-assembly of the peptide alamethicin provide a tractable model system for exploring the structural basis of charge selectivity. We synthesized covalently-linked alamethicin dimers, with amino acid substitutions at position 18 [lysine (Lys), arginine (Arg), glutamine (Gln), 2,3-diaminopropionic acid (Dpr)] in each helix, to assess the role of this position as a charge-selectivity determinant in alamethicin channels. Of the position 18 substitutions investigated, the Lys derivative exhibited the greatest degree of anion selectivity. Arg-containing channels were slightly less anion-selective than Lys. Interestingly, Dpr channels showed cation selectivity nearly equivalent to that exhibited by the neutral Gln derivative. We suggest that this result is due to a wider pore diameter that permits a greater number of counter-ions leading to enhanced charge screening and a lower effective side-chain positive charge.

Alamethicin↗
PubMed2001

A neurodynamical model for selective visual attention using oscillators.

We present a neurodynamical model to study and simulate visual search tasks experiments. The model consists of different pools of interconnected phase oscillators. Each oscillator is described by an integrate-and-fire type equation. Visual attention appears as an emergent property of the dynamic of the system, resulting from the temporal synchronization of the pools which bind the features of the searched target. The time courses observed in the psychophysical visual search experiments can be explained within a purely parallel dynamic and without assuming priority maps and serial spotlight mechanisms, as is usually done in the standard theories. The model fits also the measured activity reported for the neural responses in inferotemporal visual cortex of monkeys performing visual search tasks.

Action Potentials↗
PubMedDate not supplied

AUR Memorial Award. Two computer models for selection of optimal magnetic resonance imaging (MRI) pulse sequence timing.

Two computer modeling techniques have been developed that aid in the selection of optimal magnetic resonance imaging (MRI) pulse sequences and timing intervals for specific clinical situations. The "parameter sensitivity" technique provides a means of selecting three separate MRI scans which are individually sensitive to changes in each of the three NMR tissue parameters N, T1, and T2. The "contrast" technique allows selection of a single optimal MRI sequence using the expected changes in all three tissue parameters simultaneously. Excellent correlation is demonstrated between the models and images obtained in a normal volunteer and in a patient with multiple sclerosis. The two methods compliment each other; the parameter sensitivity method is most useful in situations where subtle changes in tissue parameters are expected, whereas the contrast method is suited to circumstances where large differences in tissue parameters are anticipated and the magnitude and direction of these changes are known.

Computers↗
PubMed1991

Mean square error of estimates of HIV prevalence and short-term AIDS projections derived by backcalculation.

We simulated multinomial AIDS incidence counts from 27 'representative' AIDS epidemics that spanned a period corresponding to previous applications of backcalculation (1 January 1977 to 1 July 1987) and assessed mean square error for several back-calculated estimators of HIV prevalence and short-term AIDS projections. Estimators were based on flexible model selection procedures that chose the best-fitting non-negatively constrained model of the infection curve from a family of possible step-function models. Selection of the best-fitting model from a family of four-step models each with a long last step of width of 4 or 4.5 years offered a favourable tradeoff between bias and variance when compared with selection from families of models with three steps or from families with a short last step. Five-step models performed as well as four-step models. Three-step models had substantially larger mean square error in some epidemic situations. Percentage root mean square error (PRMSE) for estimates of cumulative HIV prevalence as of 1 January 1985 was less than 14 per cent over a range of hypothetical epidemics of N = 50,000 infected individuals. PRMSE for short-term projections was less than 18 per cent. Estimates of cumulative HIV prevalence as of 1 July 1987 were substantially more uncertain and had a PRMSE of 33 per cent in the unfavourable case of a rapidly rising HIV epidemic. Estimates of cumulative HIV prevalence as of 1 July 1987 were positively biased in HIV epidemics with a rapidly decreasing recent HIV incidence rate and negatively biased in rapidly increasing HIV epidemics. Despite these uncertainties, we obtained useful estimates even for HIV epidemics with as few as 5000 infected individuals.

Acquired Immunodeficiency Syndrome↗
PubMed2003

Open capture-recapture models with heterogeneity: I. Cormack-Jolly-Seber model.

In open population capture-recapture studies, it is usually assumed that similar animals (e.g., of the same sex and age group) have similar survival rates and capture probabilities. These assumptions are generally perceived to be an oversimplification, and they can lead to incorrect model selection and biased parameter estimates. Allowing for individual variability in survival and capture probabilities among apparently similar animals is now becoming possible, due to advances in closed population models and improved computing power. This article presents a flexible framework of likelihood-based models which allow for individual heterogeneity in survival and capture rates. Heterogeneity is modeled using finite mixtures, which have enough flexibility of distribution shape to accommodate a wide variety of different patterns of individual variation. The models condition on the first capture of each animal, and include as a special case the Cormack-Jolly-Seber model. Model selection is done either using Akaike's information criterion or by likelihood ratio tests, making available checks of different influences on survival rates. Bias in parameter estimates is reduced by including individual heterogeneity. Model selection and bias reduction are important in population studies and for making informed management decisions.

Animals↗
PubMedDate not supplied

VSMP: a novel variable selection and modeling method based on the prediction.

The use of numerous descriptors that are indicative of molecular structure and topology is becoming more common in quantitative structure-activity relationship (QSAR). How to choose the adequate descriptors for QSAR studies is important but difficult because there are no absolute rules to govern this choice. A variety of variable selection techniques including stepwise, partial least squares/principal component analysis (PLS/PCA), neural network, and evolutionary algorithm such as genetic algorithm have been applied to this common problem. All-subsets regression (ASR) is capable of finding out the best variable subset from among a large pool. In this paper, a novel variable selection and modeling method based on the prediction, for short VSMP, has been developed. Here two controllable parameters, the interrelation coefficient between the pairs of the independent variables (r(int)) and the correlation coefficient (q(2)) obtained using the leave-one-out (LOO) cross-validation technique, are introduced into the ASR to improve its performances. This technique differs from the other variable selection procedures related to the ASR by two main features: (1) The search of various optimal subset search is controlled by the statistic q(2) or root-mean-square error (RMSEP) in the LOO cross-validation step rather than the correlation coefficient obtained in the modeling step (r(2)). (2) The searching speed of all optimal subsets is expedited by the statistic r(int) together with q(2). A comparison of the results of the VSMP applied to the Selwood data set (n = 31 compounds, m = 53 descriptors) with those obtained from alternative algorithms shows the good performance of the technique.

Journal Article↗
PubMed1995

Understanding cellular responses to toxic agents: a model for mechanism-choice in bacterial metal resistance.

Bacterial resistances to metals are heterogeneous in both their genetic and biochemical bases. Metal resistance may be chromosomally-, plasmid- or transposon-encoded, and one or more genes may be involved: at the biochemical level at least six different mechanisms are responsible for resistance. Various types of resistance mechanisms can occur singly or in combination and for a particular metal different mechanisms of resistance can occur in the same species. To understand better the diverse responses of bacteria to metal ion challenge we have constructed a qualitative model for the selection of metal resistance in bacteria. How a bacterium becomes resistant to a particular metal depends on the number and location of cellular components sensitive to the specific metal ion. Other important selective factors include the nature of the uptake systems for the metal, the role and interactions of the metal in the normal metabolism of the cell and the availability of plasmid (or transposon) encoded resistance mechanisms. The selection model presented is based on the interaction of these factors and allows predictions to be made about the evolution of metal resistance in bacterial populations. It also allows prediction of the genetic basis and of mechanisms of resistance which are in substantial agreement with those in well-documented populations. The interaction of, and selection for resistance to, toxic substances in addition to metals, such as antibiotics and toxic analogues, involve similar principles to those concerning metals. Potentially, models for selection of resistance to any substance can be derived using this approach.

Bacteria↗
PubMed2006

A novel model of selective lung ventilation to investigate the long-term effects of ventilation-induced lung injury.

Mechanical ventilation (MV) with large tidal volumes (V(T)) causes ventilator induced lung injury. Whereas immediate effects of short-term injurious ventilation are well studied, little is known about its long-term effects. We aimed to establish an animal model of selective injurious MV, permitting assessment of the long-term course of ventilation-induced lung injury. In anesthetized and instrumented rats (n = 26), laryngoscopy was performed, and one cannula for MV was placed in the left main bronchus and a second one in the trachea. Two ventilators were used to ventilate the left lung with high (20 mL/kg) and the right lung with low (4 mL/kg) V(T). In control animals, both lungs received low V(T). After 2 h of MV, animals were extubated and observed for 24 h and then killed. Left and right lungs were excised and sampled for further investigations. Survival in animals ventilated with the high V(T) was 90%. Twenty-four hours after MV, alveolar levels of humoral (tumor necrosis factor alpha, interleukin 6) and cellular (polymorphonuclear leukocytes) inflammatory markers were increased, and histological alterations were present in lungs ventilated with high V(T). A delayed decrease in PaO2 was noted 24 h after MV, with high V(T) delivered to one lung as compared with low V(T) delivered to both lungs. This animal model permits assessment of the long-term course of ventilation-induced lung injury and shows that pulmonary inflammation and histological alterations are present 24 h after unilateral injurious ventilation.

Animals↗
PubMed1995

Using sign score regression models to select variables in case-control studies.

This paper evaluates the performance of four variable selection methods suitable for case-control studies. Two of the methods are logistic regression and the rank transformed version of it which uses the ranks of the explanatory variables in place of the original observations. The third method is based on Kendall's tau b correlations. I propose a fourth method, a sign score regression model to select variables. To evaluate these four methods, I generate many data sets for a case group and a control group with the use of several different distributions and covariance matrices. I evaluate the methods on their ability to select correctly the variables related to case-control status while not selecting the unrelated variables. Using this criterion, the sign score regression method and the tau b method are more effective than the other two methods with uncorrelated or weakly correlated variables. The sign score regression method is more effective than the tau b method for all simulations that use normal variables and for some that use log-normal variables. Overall, the sign score regression method is the most effective variable selection method for data sets that have low or moderate correlations between variables.

Analysis of Variance↗
PubMed1985

The sampling distribution of linkage disequilibrium under an infinite allele model without selection.

The sampling distributions of several statistics that measure the association of alleles on gametes (linkage disequilibrium) are estimated under a two-locus neutral infinite allele model using an efficient Monte Carlo method. An often used approximation for the mean squared linkage disequilibrium is shown to be inaccurate unless the proper statistical conditioning is used. The joint distribution of linkage disequilibrium and the allele frequencies in the sample is studied. This estimated joint distribution is sufficient for obtaining an approximate maximum likelihood estimate of C = 4Nc, where N is the population size and c is the recombination rate. It has been suggested that observations of high linkage disequilibrium might be a good basis for rejecting a neutral model in favor of a model in which natural selection maintains genetic variation. It is found that a single sample of chromosomes, examined at two loci cannot provide sufficient information for such a test if C less than 10, because with C this small, very high levels of linkage disequilibrium are not unexpected under the neutral model. In samples of size 50, it is found that, even when C is as large as 50, the distribution of linkage disequilibrium conditional on the allele frequencies is substantially different from the distribution when there is no linkage between the loci. When conditioned on the number of alleles at each locus in the sample, all of the sample statistics examined are nearly independent of theta = 4N mu, where mu is the neutral mutation rate.

Alleles↗
PubMed2006

Genetic model of selective COX2 inhibition reveals novel heterodimer signaling.

Selective inhibitors of cyclooxygenase-2 (COX2) have attracted widespread media attention because of evidence of an elevated risk of cardiovascular complications in placebo-controlled trials, resulting in the market withdrawal of some members of this class. These drugs block the cyclooxygenase activity of prostaglandin H synthase-2 (PGHS2), but do not affect the associated peroxidase function. They were developed with the rationale of conserving the anti-inflammatory and analgesic actions of traditional nonsteroidal anti-inflammatory drugs (tNSAIDs) while sparing the ability of PGHS1-derived prostaglandins to afford gastric cytoprotection. PGHS1 and PGHS2 coexist in the vasculature and in macrophages, and are upregulated together in inflammatory tissues such as rheumatoid synovia and atherosclerotic plaque. They are each believed to function as homodimers. Here, we developed a new genetic mouse model of selective COX2 inhibition using a gene-targeted point mutation, resulting in a Y385F substitution. Structural modeling and biochemical assays showed the ability of PGHS1 and PGHS2 to heterodimerize and form prostaglandins. The heterodimerization of PGHS1-PGHS2 may explain how the ductus arteriosus closes normally at birth in mice expressing PGHS2 Y385F, but not in PGHS2-null mice.

Animals↗
PubMed1989

Modeling and variable selection in epidemiologic analysis.

This paper provides an overview of problems in multivariate modeling of epidemiologic data, and examines some proposed solutions. Special attention is given to the task of model selection, which involves selection of the model form, selection of the variables to enter the model, and selection of the form of these variables in the model. Several conclusions are drawn, among them: a) model and variable forms should be selected based on regression diagnostic procedures, in addition to goodness-of-fit tests; b) variable-selection algorithms in current packaged programs, such as conventional stepwise regression, can easily lead to invalid estimates and tests of effect; and c) variable selection is better approached by direct estimation of the degree of confounding produced by each variable than by significance-testing algorithms. As a general rule, before using a model to estimate effects, one should evaluate the assumptions implied by the model against both the data and prior information.

Algorithms↗
PubMed2006

Hypothesis testing and Bayesian estimation using a sigmoid Emax model applied to sparse dose-response designs.

Application of a sigmoid Emax model is described for the assessment of dose-response with designs containing a small number of doses (typically, three to six). The expanded model is a common Emax model with a power (Hill) parameter applied to dose and the ED50 parameter. The model will be evaluated following a strategy proposed by Bretz et al. (2005). The sigmoid Emax model is used to create several contrasts that have high power to detect an increasing trend from placebo. Alpha level for the hypothesis of no dose-response is controlled using multiple comparison methods applied to the p-values obtained from the contrasts. Subsequent to establishing drug activity, Bayesian methods are used to estimate the dose-response curve from the sparse dosing design. Bayesian estimation applied to the sigmoid model represents uncertainty in model selection that is missed when a single simpler model is selected from a collection of non-nested models. The goal is to base model selection on substantive knowledge and broad experience with dose-response relationships rather than criteria selected to ensure convergence of estimators. Bayesian estimation also addresses deficiencies in confidence intervals and tests derived from asymptotic-based maximum likelihood estimation when some parameters are poorly determined, which is typical for data from common dose-response designs.

Algorithms↗