The efficiency of improved ranked set sampling methods

Ranked set sampling is a desirable sampling technique where the ranking of the sample observations does not require the actual measurement. This thesis develops new ranked set sampling techniques and determines the sampling characteristics required to best estimate the required population parameters, such as mean and variance.

Diabetes MILES - Australia (Management and Impact for Long-Term Empowerment and Success) : methods and sample characteristics of a national survey of the psychological aspects of living with type 1 or type 2 diabetes in Australian adults

Background Successful management of diabetes requires attention to the behavioural, psychological and social aspects of this progressive condition. The Diabetes MILES (Management and Impact for Long-term Empowerment and Success) Study is an international collaborative. Diabetes MILES-Australia, the first Diabetes MILES initiative to be undertaken, was a national survey of adults living with type 1 or type 2 diabetes in Australia. The aim of this study was to gather data that will provide insights into how Australians manage their diabetes, the support they receive and the impact of diabetes on their lives, as well as to use the data to validate new diabetes outcome measures.

Methods The survey was designed to include a core set of self-report measures, as well as modules specific to diabetes type or management regimens. Other measures or items were included in only half of the surveys. Cognitive debriefing interviews with 20 participants ensured the survey content was relevant and easily understood. In July 2011, the survey was posted to 15,000 adults (aged 18-70 years) with type 1 or type 2 diabetes selected randomly from the National Diabetes Services Scheme (NDSS) database. An online version of the survey was advertised nationally. A total of 3,338 eligible Australians took part; most (70.4%) completed the postal survey. Respondents of both diabetes types and genders, and of all ages, were adequately represented in both the postal and online survey sub-samples. More people with type 2 diabetes than type 1 diabetes took part in Diabetes MILES-Australia (58.8% versus 41.2%). Most respondents spoke English as their main language, were married/in a de facto relationship, had at least a high school education, were occupied in paid work, had an annual household income > $AUS40,000, and lived in metropolitan areas.

Discussion A potential limitation of the study is the under-representation of respondents from culturally and linguistically diverse backgrounds (including Aboriginal and Torres Strait Islander origin). Diabetes MILES-Australia represents a major achievement in the study of diabetes in Australia, where for the first time, the focus is on psychosocial and behavioural aspects of this condition at a national level.

Effects of low- versus high-fidelity simulations on the cognitive burden and performance of entry-level paramedicine students: a mixed-methods comparison trial using eye-tracking, continuous heart rate, difficulty rating scales, video observation and inte

INTRODUCTION: High-fidelity simulation-based training is often avoided for early-stage students because of the assumption that while practicing newly learned skills, they are ill suited to processing multiple demands, which can lead to "cognitive overload" and poorer learning outcomes. We tested this assumption using a mixed-methods experimental design manipulating psychological immersion. METHODS: Thirty-nine randomly assigned first-year paramedicine students completed low- or high-environmental fidelity simulations [low-environmental fidelity simulations (LFenS) vs. high-environmental fidelity simulation (HFenS)] involving a manikin with obstructed airway (SimMan3G). Psychological immersion and cognitive burden were determined via continuous heart rate, eye tracking, self-report questionnaire (National Aeronautics and Space Administration Task Load Index), independent observation, and postsimulation interviews. Performance was assessed by successful location of obstruction and time-to-termination. RESULTS: Eye tracking confirmed that students attended to multiple, concurrent stimuli in HFenS and interviews consistently suggested that they experienced greater psychological immersion and cognitive burden than their LFenS counterparts. This was confirmed by significantly higher mean heart rate (P < 0.001) and National Aeronautics and Space Administration Task Load Index mental demand (P < 0.05). Although group allocation did not influence the proportion of students who ultimately revived the patient (58% vs. 30%, P < 0.10), the HFenS students did so significantly more quickly (P < 0.01). The LFenS students had low immersion resulting in greater assessment anxiety. CONCLUSIONS: High-environmental fidelity simulation engendered immersion and a sense of urgency in students, whereas LFenS created assessment anxiety and slower performance. We conclude that once early-stage students have learned the basics of a clinical skill, throwing them in the "deep end" of high-fidelity simulation creates significant additional cognitive burden but this has considerable educational merit.

Domain-driven classification based on multiple criteria and multiple constraint-level programming for intelligent credit scoring

Extracting knowledge from the transaction records and the personal data of credit card holders has great profit potential for the banking industry. The challenge is to detect/predict bankrupts and to keep and recruit the profitable customers. However, grouping and targeting credit card customers by traditional data-driven mining often does not directly meet the needs of the banking industry, because data-driven mining automatically generates classification outputs that are imprecise, meaningless, and beyond users' control. In this paper, we provide a novel domain-driven classification method that takes advantage of multiple criteria and multiple constraint-level programming for intelligent credit scoring. The method involves credit scoring to produce a set of customers' scores that allows the classification results actionable and controllable by human interaction during the scoring process. Domain knowledge and experts' experience parameters are built into the criteria and constraint functions of mathematical programming and the human and machine conversation is employed to generate an efficient and precise solution. Experiments based on various data sets validated the effectiveness and efficiency of the proposed methods. © 2006 IEEE.

The use of correspondence analysis to develop a scale to measure workplace morale from multi-level data

This paper focuses on the issue of comparing social groups or collectivities using measures derived from individual-level multivariate data. In this case, groups need to be differentiated such that: (a) between-group differences are maximized; (b) within-group differences are minimised; and (c) `differences' are calibrated to a scale that reflects a set indicators or observed variables.This paper demonstrates empirically how correspondence analysis can achieve this. It presents a scale of `workplace morale' derived from the responses of employees in a large sample of workplaces to questions concerning satisfaction with various facets of their job and their workplace. The scale derived through correspondence analysis is shown to achieve the three criteria described above.

Association of family environment with children's television viewing and with low level of physical activity

Objective: This study examined associations between the family environment and children's television (TV) viewing and likelihood of being low-active.
Research Methods and Procedures: In 2001, children were recruited from 19 primary schools in Melbourne, Australia. Parents completed a questionnaire about their child's TV viewing and the family environment. Children also completed a questionnaire and wore an accelerometer for 8 days. Movement counts were used to identify low-active children (lowest quartile). Data were analyzed in May 2004.
Results: The sample consisted of 878 children (mean age = 11.5 0.6 yrs). Multiple logistic regression revealed that socioeconomic status [adjusted odds ratios (AOR) = 0.4 boys], frequency families watched TV together (AOR = 2.0 boys), mothers' (AOR = 1.8 boys; AOR = 2.5 girls) and fathers' (AOR = 2.6 boys; AOR = 2.8 girls) TV viewing, and rules prohibiting TV during mealtimes (AOR = 0.6 boys; AOR = 0.6 girls) related to children watching TV 2 h/d. Variables associated with low-level physical activity included self-reported enjoyment of Internet use (AOR = 1.7 boys) and preference for watching TV (AOR = 2.3 girls), perception that mother uses computer a lot (AOR = 1.9 boys) and likes using the computer (AOR = 0.6 girls), fathers' reported computer/electronic games use (AOR = 1.7 girls), frequency families used computer together (AOR = 0.4 girls), rules that TV viewing must be supervised (AOR = 1.9 boys; AOR = 0.6 girls), and having pay TV (AOR = 0.6 boys) and electronic games at home (AOR = 2.6 boys).
Discussion: These findings suggest that the relationships between the family environment and TV viewing and low-level activity are complex and that these behaviors are distinct.

Methods of construction of OWA operators from data

This paper investigates the problem of obtaining the weights of the ordered weighted aggregation (OWA) operators from observations. The problem is formulated as a restricted least squares and uniform approximation problems. We take full advantage of the linearity of the problem. In the former case, a well known technique of non-negative least squares is used. In a case of uniform approximation, we employ a recently developed cutting angle method of global optimisation. Both presented methods give results superior to earlier approaches, and do not require complicated nonlinear constructions. Additional restrictions, such as degree of orness of the operator, can be easily introduced

An email classification model based on rough set theory

The communication via email is one of the most popular services of the Internet. Emails have brought us great convenience in our daily work and life. However, unsolicited messages or spam, flood our email boxes, which results in bandwidth, time and money wasting. To this end, this paper presents a rough set based model to classify emails into three categories - spam, no-spam and suspicious, rather than two classes (spam and non-spam) in most currently used approaches. By comparing with popular classification methods like Naive Bayes classification, the error ratio that a non-spam is discriminated to spam can be reduced using our proposed model.

Evolutionary multiobjective optimization in engineering management: an empirical study on bridge deck rehabilitation

There exist multiple objectives in engineering management such as minimum cost and maximum service capacity. Although solution methods of multiobjective optimization problems have undergone continual development over the past several decades, the methods available to date are not particularly robust, and none of them performs well on the broad classes. Because genetic algorithms work with a population of points, they can capture a number of solutions simultaneously, and easily incorporate the concept of Pareto optimal set in their optimization process. In this paper, a genetic algorithm is modified to deal with the rehabilitation planning of bridge decks at a network level by minimizing the rehabilitation cost and deterioration degree simultaneously.

Evolutionary multiobjective optimization in engineering management: an empirical application in infrastructure systems

Generally multiple objectives exist in transportation infrastructure management, such as minimum cost and maximum service capacity. Although solution methoak of multiobjective optimization problems have undergone continual development over the part several decades, the methods available to date are not particularly robust, and none of them perform well on the broad classes. Because genetic algorithms work with apopulation ofpoints, they can capture a number of solutions simultaneously, and easily incorporate the concept of a Pareto optimal set in their optimization process. In this paper, a genetic algorithm is modified to deal with an empirical application for the rehabilitation planning of bridge decks, at a network level, by minimizing the rehabilitation cost and deterioration degree simultaneously.

Methods for exploring implementation variation and local context within a cluster tandomised community intervention trial

Insignificant or modest findings in intervention trials may be attributable to poorly designed or theorised interventions, poorly implemented interventions, or inadequate evaluation methods. The pre-existing context may also account for the effects observed. A combination of qualitative and quantitative methods is outlined that will permit the determination of how context level factors might modify intervention effectiveness, within a cluster randomised community intervention trial to promote the health of mothers with new babies. The methods include written and oral narratives, key informant interviews, impact logs, and inter-organisational network analyses. Context level factors, which may affect intervention uptake, success, and sustainability are the density of inter-organisational ties within communities at the start of the intervention, the centrality of the primary care agencies expected to take a lead with the intervention, the extent of context-level adaptation of the intervention, and the amount of local resources contributed by the participating agencies. Investigation of how intervention effects are modified by context is a new methodological frontier in community intervention trial research.

Returning to work: exploring the experiences of women with acute coronary syndromes

Aims: The aim of this study was to explore the transition experiences of Australian women in resuming paid employment after an acute coronary syndrome (ACS) event. Background: Until recently cardiovascular research has focused predominantly on men but this is changing and research exploring women’s experiences of ACS has increased. Despite knowing that many women do not resume paid employment following an ACS event, little is known about the experience of those women who do, even though it is understood that returning to the previous level of employment after an ACS event is a positive outcome. Design: An exploratory qualitative approach underpinned by naturalistic inquiry was undertaken. Methods: A purposive sample of seven women who had experienced their first ACS event 12 months ago was selected. Each woman was interviewed using a semi–structured format and their interviews were transcribed verbatim. Thematic analysis of the transcript set and conceptual mapping were employed to formulate key themes. Findings: All women (mean age 52.6 years) resumed paid employment at various stages during their recovery, but reported similar transition processes. Three key themes representing this process were identified: primary motivation; influence through guidance and support; and resuming paid employment. Conclusions: Study findings revealed that these women required substantial support from family, friends and employers, with ongoing guidance from health professionals to return to paid work. The time frames for their return varied and some modified their roles within the workforce to enable them to return to paid work. However, formal cardiac rehabilitation did not appear to have a significant impact on these participants' decisions to return to work. They felt that more information about this decision may have been helpful if given at the time of cardiac rehabilitation.

Grounded theory as foundations for methods in applied ontology

Research into domain specific ontologies is difficult to treat empirically. This is because it is difficult to ground domain ontology while simultaneously being true to its guiding philosophy or theory. Further, ontology generation is often introspective and reflective or relies on experts for ontology generation. Even those relying on expert generation lack rigour and tend to be more ad-hoc. We ask how Grounded Theory can be used to generate domain specific ontologies where appropriate high level theory and suitable textual data sources are available. We are undertaking generation of a domain ontology for the discipline of information systems by applying the Grounded Theory method. Specifically we are using Roman Ingarden’s theory of scientific works to seed a coding family and adapting the method to ask relevant questions when analysing rich textual data. We have found that a guiding ontological theory, such as Ingarden’s, can be used to seed a coding family giving rise to a viable method for generating ontologies for research. This is significant because Grounded Theory may be one of the key methods for generating ontologies where substantial uniform quality text is available to the ontologist. We also present our partial analysis of information systems research.

Leukemia risk associated with low-level benzene exposure

Background: Men who were part of an Australian petroleum industry cohort had previously been found to have an excess of lympho-hematopoietic cancer. Occupational benzene exposure is a possible cause of this excess.

Methods: We conducted a case-control study of lympho-hematopoietic cancer nested within the existing cohort study to examine the role of benzene exposure. Cases identified between 1981 and 1999 (N = 79) were age-matched to 5 control subjects from the cohort. We estimated each subject's benzene exposure using occupational histories, local site-specific information, and an algorithm using Australian petroleum industry monitoring data.

Results: Matched analyses showed that the risk of leukemia was increased at cumulative exposures above 2 ppm-years and with intensity of exposure of highest exposed job over 0.8 ppm. Risk increased with higher exposures; for the 13 case-sets with greater than 8 ppm-years cumulative exposure, the odds ratio was 11.3 (95% confidence interval = 2.85-45.1). The risk of leukemia was not associated with start date or duration of employment. The association with type of workplace was explained by cumulative exposure. There is limited evidence that short-term high exposures carry more risk than the same amount of exposure spread over a longer period. The risks for acute nonlymphocytic leukemia and chronic lymphocytic leukemia were raised for the highest exposed workers. No association was found between non-Hodgkin lymphoma or multiple myeloma and benzene exposure, nor between tobacco or alcohol consumption and any of the cancers.

Conclusions: We found an excess risk of leukemia associated with cumulative benzene exposures and benzene exposure intensities that were considerably lower than reported in previous studies. No evidence was found of a threshold cumulative exposure below which there was no risk.

Weight and place: a multilevel cross-sectional survey of area-level social disadvantage and overweight/obesity in Australia

Objective: To estimate variation between small areas in adult body mass index (BMI), and assess the importance of area level socioeconomic disadvantage in predicting BMI.

Methods: We identified all census collector districts (CCDs) in the 20 innermost Local Government Areas in metropolitan Melbourne, Australia, and ranked them by the percentage of low income households (<$400/week). In all, 50 CCDs were randomly selected from the least, middle and most disadvantaged septiles of the ranked list and 4913 residents (61.4% participation rate) completed one of two surveys. Multilevel linear regression was used to estimate area level variance in BMI and the importance of area level socioeconomic disadvantage in predicting BMI.

Results: There were significant variations in BMI between CCDs for women, even after adjustment for individual and area SES (P=0.012); significant area variation was not found for men. Living in the most versus least disadvantaged areas was associated with an average difference in BMI of 1.08 kg/m² (95% CI: 0.48–1.68 kg/m²) for women, and of 0.93 kg/m² (95% CI: 0.32–1.55 kg/m²) for men. Living in the mid versus least disadvantaged areas were associated with an average difference in BMI of 0.67 kg/m² (95% CI: 0.09–1.26 kg/m²) for women, and 0.43 kg/m² for men (95% CI: -0.16–1.01).

Conclusion: These findings suggest that area disadvantage is an important predictor of adult BMI, and support the need to focus on improving local environments to reduce socioeconomic inequalities in overweight and obesity.