A renewed approach to the nonparametric analysis of replicated microarray experiments

DNA-microarrays find broad employment in biochemical research. This technology allows the monitoring of the expression levels of thousands of genes at the same time. Often, the goal of a microarray study is to find differentially expressed genes in two different types of tissue, for example normal and cancerous. Multiple hypothesis testing is a useful statistical tool for such studies. One approach using multiple hypothesis testing is nonparametric analysis for replicated microarray experiments. In this paper we present an improved version of this method. We also show how p-values are calculated for all significant genes detected with this testing procedure. All algorithms were implemented in an R-package, and instructions on it's use are included. The package can be downloaded at http://www.statistik.unidortmund.de/de/content/einrichtungen/lehrstuehle/personen/jung.html     

Uncertainty in prior elicitations: a nonparametric approach

A key task in the elicitation of expert knowledge is to constructa distribution from the finite, and usually small, number ofstatements that have been elicited from the expert. These statementstypically specify some quantiles or moments of the distribution.Such statements are not enough to identify the expert's probabilitydistribution uniquely, and the usual approach is to fit somemember of a convenient parametric family. There are two cleardeficiencies in this solution. First, the expert's beliefs areforced to fit the parametric family. Secondly, no account isthen taken of the many other possible distributions that mighthave fitted the elicited statements equally well. We presenta nonparametric approach which tackles both of these deficiencies.We also consider the issue of the imprecision in the elicitedprobability judgements.     

A Note on the Variable Kernel Estimate

By a simple example it is shown how in special cases the variable kernel estimate may not give the correct qualitative impression of the estimated density. To exclude such effects a slightly modified definition of the bandwidth of the kernels is proposed.     

A new method for estimating the size of small populations from genetic mark-recapture data

The use of non-invasive genetic sampling to estimate population size in elusive or rare species is increasing. The data generated from this sampling differ from traditional mark-recapture data in that individuals may be captured multiple times within a session or there may only be a single sampling event. To accommodate this type of data, we develop a method, named capwire, based on a simple urn model containing individuals of two capture probabilities. The method is evaluated using simulations of an urn and of a more biologically realistic system where individuals occupy space, and display heterogeneous movement and DNA deposition patterns. We also analyse a small number of real data sets. The results indicate that when the data contain capture heterogeneity the method provides estimates with small bias and good coverage, along with high accuracy and precision. Performance is not as consistent when capture rates are homogeneous and when dealing with populations substantially larger than 100. For the few real data sets where N is approximately known, capwire's estimates are very good. We compare capwire's performance to commonly used rarefaction methods and to two heterogeneity estimators in program capture: Mh-Chao and Mh-jackknife. No method works best in all situations. While less precise, the Chao estimator is very robust. We also examine how large samples should be to achieve a given level of accuracy using capwire. We conclude that capwire provides an improved way to estimate N for some DNA-based data sets.     

Loglinear models for the robust design in mark-recapture experiments

The robust design is a method for implementing a mark-recapture experiment featuring a nested sampling structure. The first level consists of primary sampling sessions; the population experiences mortality and immigration between primary sessions so that open population models apply at this level. The second level of sampling has a short mark-recapture study within each primary session. Closed population models are used at this stage to estimate the animal abundance at each primary session. This article suggests a loglinear technique to fit the robust design. Loglinear models for the analysis of mark-recapture data from closed and open populations are first reviewed. These two types of models are then combined to analyze the data from a robust design. The proposed loglinear approach to the robust design allows incorporating parameters for a heterogeneity in the capture probabilities of the units within each primary session. Temporary emigration out of the study area can also be accounted for in the loglinear framework. The analysis is relatively simple; it relies on a large Poisson regression with the vector of frequencies of the capture histories as dependent variable. An example concerned with the estimation of abundance and survival of the red-back vole in an area of southeastern Québec is presented.     

A general approach for two-stage analysis of multilevel clustered non-Gaussian data

In this article, we propose a two-stage approach to modeling multilevel clustered non-Gaussian data with sufficiently large numbers of continuous measures per cluster. Such data are common in biological and medical studies utilizing monitoring or image-processing equipment. We consider a general class of hierarchical models that generalizes the model in the global two-stage (GTS) method for nonlinear mixed effects models by using any square-root-n-consistent and asymptotically normal estimators from stage 1 as pseudodata in the stage 2 model, and by extending the stage 2 model to accommodate random effects from multiple levels of clustering. The second-stage model is a standard linear mixed effects model with normal random effects, but the cluster-specific distributions, conditional on random effects, can be non-Gaussian. This methodology provides a flexible framework for modeling not only a location parameter but also other characteristics of conditional distributions that may be of specific interest. For estimation of the population parameters, we propose a conditional restricted maximum likelihood (CREML) approach and establish the asymptotic properties of the CREML estimators. The proposed general approach is illustrated using quartiles as cluster-specific parameters estimated in the first stage, and applied to the data example from a collagen fibril development study. We demonstrate using simulations that in samples with small numbers of independent clusters, the CREML estimators may perform better than conditional maximum likelihood estimators, which are a direct extension of the estimators from the GTS method.     

A nonparametric approach to the analysis of longitudinal data via a set of level crossing problems with application to the analysis of microarray time course experiments

Here we develop a completely nonparametric method for comparing two groups on a set of longitudinal measurements. No assumptions are made about the form of the mean response function, the covariance structure or the distributional form of disturbances around the mean response function. The solution proposed here is based on the realization that every longitudinal data set can also be thought of as a collection of survival data sets where the events of interest are level crossings. The method for testing for differences in the longitudinal measurements then is as follows: for an arbitrarily large set of levels, for each subject determine the first time the subject has an upcrossing and a downcrossing for each level. For each level one then computes the log rank statistic and uses the maximum in absolute value of all these statistics as the test statistic. By permuting group labels we obtain a permutation test of the hypothesis that the joint distribution of the measurements over time does not depend on group membership. Simulations are performed to investigate the power and it is applied to the area that motivated the method-the analysis of microarrays. In this area small sample sizes, few time points and far too many genes to consider genuine gene level longitudinal modeling have created a need for a simple, model free test to screen for interesting features in the data.     

Incorporating Movement Behavior and Barriers to Improve Kernel Home Range Space Use Estimates

Abstract: Kernel-based utilization distribution (UD) estimates are powerful tools to investigate home range space use and resource selection in many vertebrate species. By ignoring local movement information provided by the serial correlation between successive locations and the constraints to movement imposed by obvious boundaries, the classical kernel method results in loosely estimated UDs that tend to overflow into never-visited areas and eventually in possibly biased estimates of space use and habitat selection. We improved biological relevance of kernel home range space use estimates by incorporating both movement (and activity) information and boundary constraints.     

A new approach to reconstruction models of dendritic branching patterns

Quantitative models for characterising the detailed branching patterns of dendritic trees aim to explain these patterns either in terms of growth models based on principles of dendritic development or reconstruction models that describe an existing structure by means of a canonical set of elementary properties of dendritic morphology, which when incorporated into an algorithmic procedure will generate samples of dendrites that are statistically indistinguishable in both canonical and emergent features from those of the original sample of real neurons. This article introduces a conceptually new approach to reconstruction modelling based on the single assumption that dendritic segments are built from sequences of units of constant diameter, and that the distribution of the lengths of units of similar diameter is independent of location within a dendritic tree. This assumption in combination with non-parametric methods for estimating univariate and multivariate probability densities leads to an algorithm that significantly reduces the number of basic parameters required to simulate dendritic morphology. It is not necessary to distinguish between stem and terminal segments or to specify daughter branch ratios or dendritic taper. The procedure of sampling probability densities conditioned on local morphological features eliminates the need, for example, to specify daughter branch ratios and dendritic taper since these emerge naturally as a consequence of the conditioning process. Thus several basic parameters of previous reconstruction algorithms become emergent parameters of the new reconstruction process. The new procedure was applied successfully to a sample of 51 interneurons from lamina II/III of the spinal dorsal horn.     

A nonparametric approach to the discrimination of two cell populations

An approach to the statistical testing of differences between two cell populations the elements of which are characterized by multivariate data is described. The approach is based on the Fisher discriminant and the Kruskal-Wallis test. The method, which is sufficient but not necessary, makes no assumptions about the normality of the data or about the equality of the covariance matrices for each population.     

A general framework of nonparametric feature selection in high-dimensional data

Nonparametric feature selection for high-dimensional data is an important and challenging problem in the fields of statistics and machine learning. Most of the existing methods for feature selection focus on parametric or additive models which may suffer from model misspecification. In this paper, we propose a new framework to perform nonparametric feature selection for both regression and classification problems. Under this framework, we learn prediction functions through empirical risk minimization over a reproducing kernel Hilbert space. The space is generated by a novel tensor product kernel, which depends on a set of parameters that determines the importance of the features. Computationally, we minimize the empirical risk with a penalty to estimate the prediction and kernel parameters simultaneously. The solution can be obtained by iteratively solving convex optimization problems. We study the theoretical property of the kernel feature space and prove the oracle selection property and Fisher consistency of our proposed method. Finally, we demonstrate the superior performance of our approach compared to existing methods via extensive simulation studies and applications to two real studies.     

A novel approach to the analysis of mass spectrally assayed stable isotope-labeling experiments

A novel approach to the analysis of mass spectrally assayed stable isotope-labeling experiments for studies of biosynthetic pathways is reported. This method determines in a mixture of product molecules, the relative number of product molecules synthesized from the stable labeled precursor pathway and those that were either present prior to the labeling period or were produced by an alternate pathway during the course of an experiment. In addition, the isotopic enrichment of the labeled atoms in the product molecules produced from the stable labeled precursor is determined. These isotopic enrichments represent the isotopic enrichment in the immediate precursors which form the product molecules and would reflect any cellular compartmentation of precursor pools. The feasibility of the method using 15NH4Cl and L-[5-15N]glutamine as precursors to study the de novo pyrimidine biosynthetic pathway in isolated rat hepatocytes is demonstrated. The results of these studies show that after incubation of rat hepatocytes with either precursor it is possible to determine the fraction of the uracil nucleotide pool that is formed by the de novo pathway during the period of exposure. The pattern of 15N labeling in the N1 and N3 positions in the uracil moiety is different for the two precursors; however, in most cases the 15N enrichment of each position remained relatively constant for each precursor with either time (15-120 min) or precursor concentration (1 to 10 mM). This method will allow the actual quantitation and isotopic enrichment of product formed by a specific biosynthetic pathway during the course of an experiment and, as such is an improvement over existing labeling techniques.     

A two-stage probabilistic approach to multiple-community similarity indices

SUMMARY: A traditional approach for assessing similarity among N (N > 2) communities is to use multiple pairwise comparisons. However, pairwise similarity indices do not completely characterize multiple-community similarity because the information shared by at least three communities is ignored. We propose a new and intuitive two-stage probabilistic approach, which leads to a general framework to simultaneously compare multiple communities based on abundance data. The approach is specifically used to extend the commonly used Morisita index and NESS (normalized expected species shared) index to the case of N communities. For comparing N communities, a profile of N- 1 indices is proposed to characterize similarity of species composition across communities. Based on sample abundance data, nearly unbiased estimators of the proposed indices and their variances are obtained. These generalized NESS and Morisita indices are applied to comparison of three size classes of plant data (seedling, saplings, and trees) within old-growth and secondary rain forest plots in Costa Rica.