A5050191159- Statistical Methods and Inference
- Statistical Methods and Bayesian Inference
- Bayesian Methods and Mixture Models
- Gaussian Processes and Bayesian Inference
- Advanced Statistical Methods and Models
- Control Systems and Identification
- Soil Geostatistics and Mapping
- Bayesian Modeling and Causal Inference
- Single-cell and spatial transcriptomics
- Image and Signal Denoising Methods
- Neural Networks and Applications
- Probabilistic and Robust Engineering Design
- Statistical Distribution Estimation and Applications
- Optimal Experimental Design Methods
- Face and Expression Recognition
- Statistical and numerical algorithms
- Spatial and Panel Data Analysis
- Gene expression and cancer classification
- Machine Learning and Algorithms
- Sparse and Compressive Sensing Techniques
- Financial Risk and Volatility Modeling
- Machine Learning and Data Classification
- Genetic and phenotypic traits in livestock
- Asthma and respiratory diseases
- Air Quality and Health Impacts
University of Technology Sydney
2015-2025
The University of Melbourne
2011-2020
Australian Research Council
2010-2020
The University of Sydney
2010-2018
ARC Centre of Excellence for Mathematical and Statistical Frontiers
2018
The University of Queensland
2018
Queensland University of Technology
2016-2017
Australian Bureau of Statistics
2016
University of Wollongong
2004-2014
Hasselt University
2011
Nonparametric regression using locally weighted least squares was first discussed by Stone and Cleveland. Recently, it shown Fan Gijbels that the local linear kernel-weighted estimator has asymptotic properties making superior, in certain senses, to Nadaraya-Watson Gasser-Muller kernel estimators. In this paper we extend their results on bias variance case of multivariate predictor variables. We are able derive leading terms for general weights matrix theory. This approach is especially...
Abstract Local least squares kernel regression provides an appealing solution to the nonparametric regression, or "scatterplot smoothing," problem, as demonstrated by Fan, for example. The practical implementation of any scatterplot smoother is greatly enhanced availability a reliable rule automatic selection smoothing parameter. In this article we apply ideas plug-in bandwidth develop strategies choosing parameter local linear estimators. Our results are applicable odd-degree polynomial...
Abstract The typical generalized linear model for a regression of response Y on predictors (X, Z) has conditional mean function based combination Z). We generalize these models to have nonparametric component, replacing the α T 0X + β 0Z by η0(α 0X) 0Z, where η0(·) is an unknown function. call partially single-index (GPLSIM). include "single-index" models, which β0 = 0. Using local methods, we propose estimates parameters (α0, β0) and obtain their asymptotic distributions. Examples...
An exact and easily computable expression for the mean integrated squared error (MISE) kernel estimator of a general normal mixture density, is given Gaussian kernels arbitrary order. This provides powerful new way understanding density estimation which complements usual tools simulation asymptotic analysis. The family densities very flexible formulae derived allow simple analysis wide variety shapes. A number applications this method giving important insights into are presented. Among these...
Variational approximations facilitate approximate inference for the parameters in complex statistical models and provide fast, deterministic alternatives to Monte Carlo methods. However, much of contemporary literature on variational is Computer Science rather than Statistics, uses terminology, notation, examples from former field. In this article we explain approximation terms. particular, illustrate ideas using that are familiar statisticians.
Twitching motility-mediated biofilm expansion is a complex, multicellular behavior that enables the active colonization of surfaces by many species bacteria. In this study we have explored emergence intricate network patterns interconnected trails form in actively expanding biofilms Pseudomonas aeruginosa. We used high-resolution, phase-contrast time-lapse microscopy and developed sophisticated computer vision algorithms to track analyze individual cell movements during P. aeruginosa...
A family of prior distributions for covariance matrices is studied. Members the possess attractive property all standard deviation and correlation parameters being marginally noninformative particular hyperparameter choices. Moreover, quite simple and, approximate Bayesian inference techniques such as Markov chain Monte Carlo mean field variational Bayes, has tractability on par with Inverse-Wishart conjugate distributions. simulation study shows that new can lead to more accurate sparse...
Abstract We investigate the extension of nonparametric regression technique local polynomial fitting with a kernel weight to generalized linear models and quasi-likelihood contexts. In ordinary case, has been seen have several appealing features in terms intuitive mathematical simplicity. One noteworthy feature is better performance near boundaries compared traditional estimators. These properties are shown carry over model settings. also derive asymptotic distributions proposed class...
Abstract The basic kernel density estimator in one dimension has a single smoothing parameter, usually referred to as the bandwidth. For higher dimensions, however, there are several options for parameterization of estimator. bivariate case, can be between and three independent parameters estimator, which leads flexibility versus complexity trade-off when using this practice. In article performances different possible parameterizations compared, both asymptotic exact mean integrated squared...
Penalized splines can be viewed as BLUPs in a mixed model framework, which allows the use of software for smoothing. Thus, originally developed Bayesian analysis models used penalized spline regression. inference nonparametric enjoys flexibility and exact provided by inferential machinery. This paper provides simple, yet comprehensive, set programs implementation WinBUGS. Good mixing properties MCMC chains are obtained using low-rank thin-plate splines, while simulation times per iteration...
Summary A study into geographical variability of reproductive health outcomes (e.g. birth weight) in Upper Cape Cod, Massachusetts, USA, benefits from geostatistical mapping or kriging. However, also observed are some continuous covariates maternal age) that exhibit pronounced non-linear relationships with the response variable. To account for such effects properly we merge kriging additive models to obtain what call geoadditive models. The merging becomes effortless by expressing both as...
Abstract For the density estimation problem global window width kernel estimator does not perform well when underlying has features that require different amounts of smoothing at locations. In this article we propose to transform data with intention a is more appropriate for transformed data. The estimate original “back-transform” by change variables data's density. We explore choosing transformation from suitable parametric families. Data-based selection rules choice transformations and are...
Abstract The most important parameter of a histogram is the bin width because it controls tradeoff between presenting picture with too much detail ("undersmoothing") or little ("oversmoothing") respect to true distribution. Despite this importance there has been surprisingly research into estimation "optimal" width. Default widths in common statistical packages are, at least for large samples, quite far from optimal Rules proposed by, example, Scott lead better sample performance histogram,...
Semiparametric regression is a fusion between parametric and nonparametric that integrates low-rank penalized splines, mixed model hierarchical Bayesian methodology – thus allowing more streamlined handling of longitudinal spatial correlation. We review progress in the field over five-year period 2003 2007. find semiparametric to be vibrant with substantial involvement activity, continual enhancement widespread application.
Summary An exposition on the use of O'Sullivan penalized splines in contemporary semiparametric regression, including mixed model and Bayesian formulations, is presented. are similar to P‐splines, but have advantage being a direct generalization smoothing splines. Exact expressions for penalty matrix obtained. Comparisons between two types reveal that more closely mimic natural boundary behaviour Implementation modern computing environments such as Matlab , r bugs discussed.
SUMMARY We describe semiparametric estimation and inference in a logistic regression model with measurement error the predictors. The particular consists of primary data set which only response Y fallible surrogate W true predictor X are observed, plus smaller validation for (Y, X, W) observed. Except underlying assumption predictor, no parametric distributional is made about or its surrogate. develop parameter estimate asymptotically normally distributed computationally feasible. relies on...
We investigate kernel estimators of multivariate density derivative functions using general (or unconstrained) bandwidth matrix selectors.These have been relatively less well researched than their estimator analogues.A major obstacle for progress has the intractability analysis when treating higher order derivatives.With an alternative vectorization these derivatives, mathematical intractabilities are surmounted in elegant and unified framework.The finite sample asymptotic squared errors...
We develop strategies for mean field variational Bayes approximate inference Bayesian hierarchical models containing elaborate distributions. loosely define distributions to be those having more complicated forms compared with common such as in the Normal and Gamma families. Examples are Asymmetric Laplace, Skew Generalized Extreme Value Such suffer from difficulty that parameter updates do not admit closed form solutions. circumvent this problem through a combination of (a) specially...
The conditional variance function in a heteroscedastic, nonparametric regression model is estimated by linear smoothing of squared residuals. Attention focused on local polynomial smoothers. Both the mean and functions are assumed to be smooth, but neither parametric family. biasing effect preliminary estimation studied, degrees-of-freedom correction bias proposed. corrected method shown adaptive sense that can with same asymptotic as if were known. A proposal made for using standard...
Abstract We present a simple semiparametric model for fitting subject‐specific curves longitudinal data. Individual are modelled as penalized splines with random coefficients. This has mixed representation, and it is easily implemented in standard statistical software. conduct an analysis of the long‐term effect radiation therapy on height children suffering from acute lymphoblastic leukaemia using framework effects models. The revealed significant differences between therapies showed that...