Hierarchical models for 2D presence/absence data having ambiguous zeroes: With a biogeographical case study on dingo behaviour

This dissertation is primarily an applied statistical modelling investigation, motivated by a case study comprising real data and real questions. Theoretical questions on modelling and computation of normalization constants arose from pursuit of these data analytic questions. The essence of the thesis can be described as follows. Consider binary data observed on a two-dimensional lattice. A common problem with such data is the ambiguity of zeroes recorded. These may represent zero response given some threshold (presence) or that the threshold has not been triggered (absence). Suppose that the researcher wishes to estimate the effects of covariates on the binary responses, whilst taking into account underlying spatial variation, which is itself of some interest. This situation arises in many contexts and the dingo, cypress and toad case studies described in the motivation chapter are examples of this. Two main approaches to modelling and inference are investigated in this thesis. The first is frequentist and based on generalized linear models, with spatial variation modelled by using a block structure or by smoothing the residuals spatially. The EM algorithm can be used to obtain point estimates, coupled with bootstrapping or asymptotic MLE estimates for standard errors. The second approach is Bayesian and based on a three- or four-tier hierarchical model, comprising a logistic regression with covariates for the data layer, a binary Markov Random field (MRF) for the underlying spatial process, and suitable priors for parameters in these main models. The three-parameter autologistic model is a particular MRF of interest. Markov chain Monte Carlo (MCMC) methods comprising hybrid Metropolis/Gibbs samplers is suitable for computation in this situation. Model performance can be gauged by MCMC diagnostics. Model choice can be assessed by incorporating another tier in the modelling hierarchy. This requires evaluation of a normalization constant, a notoriously difficult problem. Difficulty with estimating the normalization constant for the MRF can be overcome by using a path integral approach, although this is a highly computationally intensive method. Different methods of estimating ratios of normalization constants (N Cs) are investigated, including importance sampling Monte Carlo (ISMC), dependent Monte Carlo based on MCMC simulations (MCMC), and reverse logistic regression (RLR). I develop an idea present though not fully developed in the literature, and propose the Integrated mean canonical statistic (IMCS) method for estimating log NC ratios for binary MRFs. The IMCS method falls within the framework of the newly identified path sampling methods of Gelman & Meng (1998) and outperforms ISMC, MCMC and RLR. It also does not rely on simplifying assumptions, such as ignoring spatio-temporal dependence in the process. A thorough investigation is made of the application of IMCS to the three-parameter Autologistic model. This work introduces background computations required for the full implementation of the four-tier model in Chapter 7. Two different extensions of the three-tier model to a four-tier version are investigated. The first extension incorporates temporal dependence in the underlying spatio-temporal process. The second extensions allows the successes and failures in the data layer to depend on time. The MCMC computational method is extended to incorporate the extra layer. A major contribution of the thesis is the development of a fully Bayesian approach to inference for these hierarchical models for the first time. Note: The author of this thesis has agreed to make it open access but invites people downloading the thesis to send her an email via the 'Contact Author' function.

A Python package for Bayesian estimation using Markov Chain Monte Carlo

Markov chain Monte Carlo (MCMC) estimation provides a solution to the complex integration problems that are faced in the Bayesian analysis of statistical problems. The implementation of MCMC algorithms is, however, code intensive and time consuming. We have developed a Python package, which is called PyMCMC, that aids in the construction of MCMC samplers and helps to substantially reduce the likelihood of coding error, as well as aid in the minimisation of repetitive code. PyMCMC contains classes for Gibbs, Metropolis Hastings, independent Metropolis Hastings, random walk Metropolis Hastings, orientational bias Monte Carlo and slice samplers as well as specific modules for common models such as a module for Bayesian regression analysis. PyMCMC is straightforward to optimise, taking advantage of the Python libraries Numpy and Scipy, as well as being readily extensible with C or Fortran.

Does the social inclusion discourse add value?

The election of a national Labor Government in 2007 saw ‘social inclusion’ emerge as Australia’s overarching social policy agenda. Being ‘included’ has since been defined as being able to ‘have the resources, opportunities and capabilities needed to learn, work, engage and have a voice’. Various researchers have adopted the social inclusion concept to construct a multi-dimensional framework for measuring disadvantage, beyond poverty alleviation. This research program has enabled various forms of statistical modelling based on some agreement about what it means to be ‘included’ in society. At the same time it is acknowledged that social inclusion remains open and contestable and can be used in the name of both progressive and more punitive programs and policies. This ambiguity raises questions about whether the social inclusion framework, as it is presently defined, has the potential to be a progressive and transformative discourse. In this paper we examine whether the Australian social inclusion agenda has the capacity to address social inequality in a meaningful way, concluding with a discussion about the need to understand social inequality and social disadvantage in relational terms.

A safe system-based approach to selection of clear zones, safety barriers and other roadside treatments

On average, 560 fatal run-off-road crashes occur annually in Australia and 135 in New Zealand. In addition, there are more than 14,000 run-off-road crashes causing injuries each year across both countries. In rural areas, run-off-road casualty crashes constitute 50-60% of all casualty crashes. Their severity is particularly high with more than half of those involved sustaining fatal or serious injuries. This paper reviews the existing approach to roadside hazard risk assessment, selection of clear zones and hazard treatments. It proposes a modified approach to roadside safety evaluation and management. It is a methodology based on statistical modelling of run-off-road casualty crashes, and application of locally developed crash modification factors and severity indices. Clear zones, safety barriers and other roadside design/treatment options are evaluated with a view to minimise fatal and serious injuries – the key Safe System objective. The paper concludes with a practical demonstration of the proposed approach. The paper is based on findings from a four-year Austroads research project into improving roadside safety in the Safe System context.

Tangled Webs : Using Bayesian Networks in the Fight Against Infection

Bayesian networks (BNs) provide a statistical modelling framework which is ideally suited for modelling the many factors and components of complex problems such as healthcare-acquired infections. The methicillin-resistant Staphylococcus aureus (MRSA) organism is particularly troublesome since it is resistant to standard treatments for Staph infections. Overcrowding and understa�ng are believed to increase infection transmission rates and also to inhibit the effectiveness of disease control measures. Clearly the mechanisms behind MRSA transmission and containment are very complicated and control strategies may only be e�ective when used in combination. BNs are growing in popularity in general and in medical sciences in particular. A recent Current Content search of the number of published BN journal articles showed a fi�ve fold increase in general and a six fold increase in medical and veterinary science from 2000 to 2009. This chapter introduces the reader to Bayesian network (BN) modelling and an iterative modelling approach to build and test the BN created to investigate the possible role of high bed occupancy on transmission of MRSA while simultaneously taking into account other risk factors.

Ignition delay of bio-fuels in a common-rail compression ignition engine

The utility of a novel technique for determining the ignition delay in a compression ignition engine has been shown. This method utilises statistical modelling in the Bayesian paradigm to accurately resolve the start of combustion from a band-pass in-cylinder pressure signal. Applied to neat diesel and six biofuels, including four fractionations of palm oil of varying carbon chain length and degree of unsaturation, the relationships between ignition delay, cetane number and oxygen content have been explored. It is noted that the expected negative relationship between ignition delay and cetane number held, as did the positive relationship between ignition delay and oxygen content. The degree of unsaturation was also identified as a potential factor influencing the ignition delay.

Approximate Bayesian computation using auxiliary model based estimates

We present a novel approach for developing summary statistics for use in approximate Bayesian computation (ABC) algorithms using indirect infer- ence. We embed this approach within a sequential Monte Carlo algorithm that is completely adaptive. This methodological development was motivated by an application involving data on macroparasite population evolution modelled with a trivariate Markov process. The main objective of the analysis is to compare inferences on the Markov process when considering two di®erent indirect mod- els. The two indirect models are based on a Beta-Binomial model and a three component mixture of Binomials, with the former providing a better ¯t to the observed data.