Survey data are often subject to a number of measurement errors. The measurement error associated with a multinomial variable is called a misclassification error. In this dissertation we study such errors when the outcome is binary. It is known that ignoring such misclassification errors may affect the parameter estimates, see for example Hausman, Abrevaya and Scott-Morton (1998). However, previous studies showed that robust estimation of the parameters is achievable if we take misclassification into account. There are many attempts to do so in the literature and the major problem in implementing them is to avoid poor or fragile identifiability of the misclassification probabilities. Generally we restrict these parameters by imposing prior information on them. Such prior constraints on the parameters are simple to impose within a Bayesian framework. Hence we consider a Bayesian logistic regression model that takes into account the misclassification of the dependent variable. A very convenient way to implement such a Bayesian analysis is to estimate the hierarchical model using the WinBUGS software package developed by the MRC biostatistics group, Institute of Public Health, at Cambridge University. WinGUGS allows us to estimate the posterior distributions of all the parameters using relatively little programming and once the program is written it is trivial to change the link function, for example from logit to probit. If we wish to have more control over the sampling scheme or to deal with more complex models, then we propose a data augmentation approach using the Metropolis-Hastings algorithm within a Gibbs sampling framework. The sampling scheme can be made more efficient by using a one-step Newton-Raphson algorithm to form the Metropolis-Hastings proposal. Results from empirically analyzing real data and from the simulation studies suggest that if suitable priors are specified for the misclassification parameters and the regression parameters, then logistic regression allowing for misclassification results in better estimators than the estimators that do not take misclassification into account.

This thesis describes the development of a decision support tool for unplanned maintenance of aircraft at ramp time during airport operations. Ramp time is the time between an aircraft arrival and its next departure. Clearance of an aircraft for flight is controlled by aviation regulations. Therefore decisions regarding maintenance are taken by engineers who have to comply with the regulations that are governed outside the organizational structure of the airline. Unplanned maintenance also often disrupts the normal operational scheduling and leads to significant costs. Therefore, the decision support tool must include the relevant aviation regulations, be capable of rescheduling to minimise disruption and be able to optimise solutions based on cost. In this project an aircraft schedule is used to demonstrate the procedures. An assumed fleet of six airplanes fly between three cities. Consultation with aviation experts ensured the size of the fleet and operations are realistic. A regulation database was developed based on the Master Minimum Equipment List (MMEL) for the aircraft, and a computer programme was developed to provide different options that comply with the regulations and take into account scheduling disruption and costs. In certain cases the regulations allow an aircraft to fly with some components inoperable so long as backup systems can perform the tasks. It is possible then to postpone the maintenance until the aircraft arrives at a properly equipped airport, or until a longer scheduled stopover reduces the disruption to operations. To address the engineering aspects of the project, maintenance of a single component that appears in the MMEL for the chosen aircraft is considered. To plan maintenance following a failure, the cause of the failure needs to be identified. Only then can the resources and time required to repair the defect be defined. The programme validation has confirmed it is able to balance different aspects of decisions related to unplanned aircraft ramp maintenance. Although the programme is based on an assumed fleet operation, the structure of the programme will allow it to be applied to other fleet and route configurations.

This thesis investigates how an autonomous reinforcement learning agent can improve on an approximate solution by augmenting it with a small patch, which overrides the approximate solution at certain states of the problem. In reinforcement learning, many approximate solutions are smaller and easier to produce than ???flat??? solutions that maintain distinct parameters for each fully enumerated state, but the best solution within the constraints of the approximation may fall well short of global optimality. This thesis proposes that the remaining gap to global optimality can be efficiently minimised by learning a small patch over the approximate solution. In order to improve the agent???s behaviour, algorithms are presented for learning the overriding patch. The patch is grown around particular regions of the problem where the approximate solution is found to be deficient. Two heuristic strategies are proposed for concentrating resources to those areas where inaccuracies in the approximate solution are most costly, drawing a compromise between solution quality and storage requirements. Patching also handles problems with continuous state variables, by two alternative methods: Kuhn triangulation over a fixed discretisation and nearest neighbour interpolation with a variable discretisation. As well as improving the agent???s behaviour, patching is also applied to the agent???s model of the environment. Inaccuracies in the agent???s model of the world are detected by statistical testing, using a selective sampling strategy to limit storage requirements for collecting data. The patching algorithms are demonstrated in several problem domains, illustrating the effectiveness of patching under a wide range of conditions. A scenario drawn from a real-time strategy game demonstrates the ability of patching to handle large complex tasks. These contributions combine to form a general framework for patching over approximate solutions in reinforcement learning. Complex problems cannot be solved by brute force alone, and some form of approximation is necessary to handle large problems. However, this does not mean that the limitations of approximate solutions must be accepted without question. Patching demonstrates one way in which an agent can leverage approximation techniques without losing the ability to handle fine yet important details.

Over the past 50 years, a great deal of research has conceptualised and modelled consumer decision-making as a single-or two-stage decision process. Today, the decision complexity has increased and consumers need to filter out a large amount of information prior to the final choice decision. This poses a challenge for marketing modellers to develop decision models that are more representative of real-world decision-making. An important rationale for the present study is to improve our understanding of consumer decision-making by providing empirical evidence that consumer decision-making may go beyond a single-or two-stage structure. This thesis aims to provide an insightful view of consumer decision-making, which may help marketers to develop and reinforce marketing programs to address consumer needs and hence increase profits, with knowledge of the types of decisions made and how decisions are made at different stages of the decision process. The literature review identified single-and two-stage decision models. Data analysis did not fully support this conceptualisation. An empirical exploration of consumer decision-making for a durable product revealed that the existing literature is limited in scope and predictability as they failed to capture multi-phase decision processes, which accounted for approximately one-half of consumer decisions. Empirical evidence was found suggesting that consumers seldom use a single strategy throughout the decision process. Consumer heterogeneity was also evident in this research as different consumers approached the same decision task with different processes and outcomes. Finally, this research identified those aspects of decision processes that have not been captured by the literature-based decision strategies. This research suggests that consumer decisions are more contingent than previously conceived in a single-and two-stage model. This research recommends that marketers should reconsider their understanding of consumer decision-making and bear in mind that one marketing strategy does not fit all customers. Marketers need to develop marketing strategies to address the entire decision process instead of focusing only on the decision outcome. By identifying different decision paths that are used by consumers, marketers can effectively segment the market; marketers can also benchmark consumers' perceptions of their performance on the important attributes against competitors to ensure that their product/brand is not eliminated prior to the selection from within the choice set. Future research requires us to understand how consumer differences interact with the decision environment to influence decision processes and outcomes. To do so, researchers must adopt a multi-phase perspective.

The aim of this thesis is to describe how men with localised prostate cancer make decisions about their medical treatments, to describe the psychological and decision-related adjustment of these men over time, and to identify what variables predict decision-related adjustment. Chapter 1 reviews the medical context of localised prostate cancer and factors that influence men's decision making in this context. It is concluded that owing to ongoing uncertainty about the optimal medical treatment for this cancer and the substantial negative quality of life effects of treatments, how men make decisions in this context is an important research question. Further, although men with prostate cancer are high seekers of medical treatment information, knowledge about how men use such information and actually make this treatment decision is limited. Chapter 2 discusses research approaches currently applied to patient decision making: first, a social interaction approach encompassing the interaction between the patient and their physician and the social context influencing this interaction (Charles, Gafni, & Whelan, 1999); and second, normative decision theory (Shafir & Tversky, 1992; von Neumann & Morgenstern, 1947). The Heuristic-Systematic Processing model (Chaiken, 1980) is then proposed as a theoretical framework for investigating patient decision making that includes both systematic and non-systematic decision strategies. Chapter 3 reviews applied decision research in cancer, and presents an overview of research findings regarding patients' preferences for involvement in decision making, the relationship between decisional involvement and psychological adjustment, and decisional support interventions. Research on adjustment to cancer is discussed and the need for further research about men's psychological and decision-related adjustment after localised prostate cancer is identified. Finally, a multivariate analysis of decision-related adjustment for men with localised prostate cancer based on the stress and coping framework of Lazarus and Folkman (1984) is proposed. Chapter 4 describes Study 1 that was an experiment to investigate the utility of the Heuristic-Systematic Processing model (HSM) in explaining low desire for involvement in decision making about prostate cancer treatments as an example of use of the expert opinion heuristic. Using a hypothetical decision scenario about localised prostate cancer it was found that a low desire for involvement in decision making by men was predicted by a high belief in powerful others controlling health, a low belief in the self being responsible for good health, a high preference for black and white thinking, and a lower level of education. This study provides preliminary support for use of the HSM in this context and for the conceptualisation of decision deferral as the expert opinion heuristic. Chapter 5 introduces and describes the method of Study 2 that was a descriptive, prospective study of men's decision making after an actual diagnosis of localised prostate cancer. This method allowed for an analysis of men's decision making that includes both systematic and non-systematic processes, and for further investigation of the utility of the HSM in explaining decision behaviour. In addition, a multivariate approach was used to describe men's physical, psychological and decision-related adjustment over time, and to identify psychological predictors of decision-related adjustment. Chapters 6 describes men's use of systematic processing as limited and the use of non-systematic processes, such as lay beliefs and heuristics, as pervasive. It is concluded that patients do not utilise information about medical treatments in a comprehensive or systematic way when making treatment decisions and that patients' decision making is biased by their prior beliefs about cancer and health. A framework is outlined to demonstrate how the results of Study 1 and 2 support the application of the HSM to decision making about prostate cancer with particular reference to the role of beliefs about the physician, health locus of control, and uncertainty about the treatment decision in influencing decision strategies. Chapter 7 describes men's physical, psychological and decision-related adjustment over time, and concludes that decision-related distress is high but psychological distress in general is low. Decisional conflict at diagnosis and at twelve months concurrently, and at two months prospectively, was predicted by dispositional optimism; and this effect was mediated by the man's proximal cognitive appraisal of the impact of the cancer. It is concluded that decisional conflict is a person's cognitive judgment of the treatment decision that is generated by similar processes to that of the psychological distress that follows a cancer diagnosis. Conclusions and implications of these studies for future research in this area are summarised in Chapter 8.

Many midwives continue to use Continuous Electronic Foetal Monitoring (CEFM) on low risk women in labour, despite overwhelming clinical evidence that it is unnecessary. The use of CEFM on low risk labouring women has been linked to rising rates of medical intervention during labour and birth with no improvement in long term neonatal outcomes. This study examined the decision-making processes of midwives who used CEFM on low risk labouring women. Whilst a number of previous studies have examined various aspects of CEFM, none specific to midwives’ decision-making and CEFM on low risk labouring women. This study contributes to the literature in this specific area. The theoretical origins of Symbolic Interactionism and Grounded Theory (GT) methods underpin this study. SI, a sociological theory that emphasises meaning in human interactions and behaviours is used in this study to focus on the behaviours and interactions of five midwives’when deciding to use CEFM on low risk labouring women. Primary data were collected by conducting unstructured interviews and systematic analysis was undertaken using GT methods to generate a substantive theory of: Midwives’ CEFM decision-making despite evidence based guidelines. The midwives made the decision that led to CEFM at two key points in the woman’s labour care. Firstly, during the initial assessment of the woman and foetus, some midwives decided to use a baseline CTG rather than intermittent auscultation (IA). Secondly, following initial assessment, the midwives made an individualised assessment and decided whether to use CEFM as the method to monitor the foetus during labour. Trust was identified as the core variable, having a profound effect on the midwives’ decision-making at these two points. Another significant factor that impacted on decision-making was staff workload. Recommendations relating to these findings promote that labouring women be central and intimately involved in decisions about foetal monitoring. Workplace reforms, such as the introduction of midwifery led models of care for women within a community setting are recommended to address professional trust and workload issues. Through the implementation of these recommendations it is expected that midwives will embrace the notion of woman centred care and that the unnecessary use of CEFM on low risk labouring women will be reduced.

One method chosen to achieve efficiency and effectiveness in the health and education sectors is a decentralized voluntary board system. In Canada, boards are delegated specific administrative responsibilities but much of the power for funding and policy development is still held centrally by the provincial governments. Depending on the specifics of the debate, this structure creates either a dependent of an interdependent relationship between the boards and the provincial government that created them. What the relationship should be is unclear because the responsibilities associated with governance are not well defined in the literature. Given this reality, board members continue to accept the responsibility to account to the provincial governments and represent many stakeholders.

This thesis explains to statisticians what graphical models are and how to use them for statistical inference; in particular, how to use decomposable graphical models for efficient inference in covariance selection and multivariate regression problems. The first aim of the thesis is to show that decomposable graphical models are worth using within a Bayesian framework. The second aim is to make the techniques of graphical models fully accessible to statisticians. To achieve these aims the thesis makes a number of statistical contributions. First, it proposes a new prior for decomposable graphs and a simulation methodology for estimating this prior. Second, it proposes a number of Markov chain Monte Carlo sampling schemes based on graphical techniques. The thesis also presents some new graphical results, and some existing results are reproved to make them more readily understood. Appendix 8.1 contains all the programs written to carry out the inference discussed in the thesis, together with both a summary of the theory on which they are based and a line by line description of how each routine works.

PhD / This thesis is concerned with planning and acting under uncertainty in partially-observable continuous domains. In particular, it focusses on the problem of mobile robot navigation given a known map. The dominant paradigm for robot localisation is to use Bayesian estimation to maintain a probability distribution over possible robot poses. In contrast, control algorithms often base their decisions on the assumption that a single state, such as the mode of this distribution, is correct. In scenarios involving significant uncertainty, this can lead to serious control errors. It is generally agreed that the reliability of navigation in uncertain environments would be greatly improved by the ability to consider the entire distribution when acting, rather than the single most likely state. The framework adopted in this thesis for modelling navigation problems mathematically is the Partially Observable Markov Decision Process (POMDP). An exact solution to a POMDP problem provides the optimal balance between reward-seeking behaviour and information-seeking behaviour, in the presence of sensor and actuation noise. Unfortunately, previous exact and approximate solution methods have had difficulty scaling to real applications. The contribution of this thesis is the formulation of an approach to planning in the space of continuous parameterised approximations to probability distributions. Theoretical and practical results are presented which show that, when compared with similar methods from the literature, this approach is capable of scaling to larger and more realistic problems. In order to apply the solution algorithm to real-world problems, a number of novel improvements are proposed. Specifically, Monte Carlo methods are employed to estimate distributions over future parameterised beliefs, improving planning accuracy without a loss of efficiency. Conditional independence assumptions are exploited to simplify the problem, reducing computational requirements. Scalability is further increased by focussing computation on likely beliefs, using metric indexing structures for efficient function approximation. Local online planning is incorporated to assist global offline planning, allowing the precision of the latter to be decreased without adversely affecting solution quality. Finally, the algorithm is implemented and demonstrated during real-time control of a mobile robot in a challenging navigation task. We argue that this task is substantially more challenging and realistic than previous problems to which POMDP solution methods have been applied. Results show that POMDP planning, which considers the evolution of the entire probability distribution over robot poses, produces significantly more robust behaviour when compared with a heuristic planner which considers only the most likely states and outcomes.

University of Technology, Sydney. School of Management, Faculty of Business. / This study investigated how selected Australian universities evaluated and adopted various learning management systems in their teaching and learning programs, given claims of uncritical evaluation, problems and cautions in the Australian (1998: 13; Brabazon, 2002; Yetton, Forster, Hewson, Hughes, Johnston, Nightingale, Page-Hanify, Vitale and Wills, 1997) and North American (Berg, 2002; Noble, 1998b) higher education literatures. Ironically, universities charge large amounts of money teaching their students to develop competence in critical analysis, yet some studies have claimed that they were deficient in critically analysing their own decisions (Brabazon, 2002; Yetton et al., 1997). This important question has received little attention in the higher education literature, despite the high visibility and costs of these decisions. Although limited theoretical explanations have been proposed by various researchers, such as Yetton et al. (1997) and Brabazon (2002), these matters have not been the subject of published empirical research to date. A grounded theory methodological framework, validated by the insights of institutional theory, was employed throughout to promote broader sociological explanations than other studies constrained by functionalist theoretical frameworks (Yetton et al., 1997). Qualitative case studies utilising semi-structured interviews and document analysis were conducted at three Australian universities. The findings of this analysis were written up in three case study narratives and an analytic cross-case analysis. Semi-structured interviews and document analysis at the field level were undertaken as an additional source of data to verify emergent grounded theory. A grounded theory of uncritical decision making (Figure 57) was ultimately developed in response to this study’s research problem. The core category around which this model was developed (‘falling behind’) appeared in all three cases, in interviews with experts from the Australian higher education sector, and was also found in both the Australian and overseas higher education literatures. This grounded theory also represents a minor contribution to the institutional theory literature as a new institutional change process model which links the activities of key individuals with broader field developments, and integrates the constructive and reproductive assumptions of old and new institutional theory.