Hazard based models for freeway traffic incident duration

Assessing and prioritising cost-effective strategies to mitigate the impacts of traffic incidents and accidents on non-recurrent congestion on major roads represents a significant challenge for road network managers. This research examines the influence of numerous factors associated with incidents of various types on their duration. It presents a comprehensive traffic incident data mining and analysis by developing an incident duration model based on twelve months of incident data obtained from the Australian freeway network. Parametric accelerated failure time (AFT) survival models of incident duration were developed, including log-logistic, lognormal, and Weibul-considering both fixed and random parameters, as well as a Weibull model with gamma heterogeneity. The Weibull AFT models with random parameters were appropriate for modelling incident duration arising from crashes and hazards. A Weibull model with gamma heterogeneity was most suitable for modelling incident duration of stationary vehicles. Significant variables affecting incident duration include characteristics of the incidents (severity, type, towing requirements, etc.), and location, time of day, and traffic characteristics of the incident. Moreover, the findings reveal no significant effects of infrastructure and weather on incident duration. A significant and unique contribution of this paper is that the durations of each type of incident are uniquely different and respond to different factors. The results of this study are useful for traffic incident management agencies to implement strategies to reduce incident duration, leading to reduced congestion, secondary incidents, and the associated human and economic losses.

Some extensions to representation and encoding of structure in models of distributional semantics

This thesis makes several contributions towards improved methods for encoding structure in computational models of word meaning. New methods are proposed and evaluated which address the requirement of being able to easily encode linguistic structural features within a computational representation while retaining the ability to scale to large volumes of textual data. Various methods are implemented and evaluated on a range of evaluation tasks to demonstrate the effectiveness of the proposed methods.

Random Breath Testing in Australia: Is there an Optimum Level of Intensity?

Background: Random Breath Testing (RBT) is the main drink driving law enforcement tool used throughout Australia. International comparative research considers Australia to have the most successful RBT program compared to other countries in terms of crash reductions (Erke, Goldenbeld, & Vaa, 2009). This success is attributed to the programs high intensity (Erke et al., 2009). Our review of the extant literature suggests that there is no research evidence that indicates an optimal level of alcohol breath testing. That is, we suggest that no research exists to guide policy regarding whether or not there is a point at which alcohol related crashes reach a point of diminishing returns as a result of either saturated or targeted RBT testing. Aims: In this paper we first provide an examination of RBTs and alcohol related crashes across Australian jurisdictions. We then address the question of whether or not an optimal level of random breath testing exists by examining the relationship between the number of RBTs conducted and the occurrence of alcohol-related crashes over time, across all Australian states. Method: To examine the association between RBT rates and alcohol related crashes and to assess whether an optimal ratio of RBT tests per licenced drivers can be determined we draw on three administrative data sources form each jurisdiction. Where possible data collected spans January 1st 2000 to September 30th 2012. The RBT administrative dataset includes the number of Random Breath Tests (RBTs) conducted per month. The traffic crash administrative dataset contains aggregated monthly count of the number of traffic crashes where an individual’s recorded BAC reaches or exceeds 0.05g/ml of alcohol in blood. The licenced driver data were the monthly number of registered licenced drivers spanning January 2000 to December 2011. Results: The data highlights that the Australian story does not reflective of all States and territories. The stable RBT to licenced driver ratio in Queensland (of 1:1) suggests a stable rate of alcohol related crash data of 5.5 per 100,000 licenced drivers. Yet, in South Australia were a relative stable rate of RBT to licenced driver ratio of 1:2 is maintained the rate of alcohol related traffic crashes is substantially less at 3.7 per 100,000. We use joinpoint regression techniques and varying regression models to fit the data and compare the different patterns between jurisdictions. Discussion: The results of this study provide an updated review and evaluation of RBTs conducted in Australia and examines the association between RBTs and alcohol related traffic crashes. We also present an evidence base to guide policy decisions for RBT operations.

Identify disorders in health records using Conditional Random Fields and Metamap: AEHRC at ShARe/CLEF 2013 eHealth Evaluation Lab Task 1

The Australian e-Health Research Centre (AEHRC) recently participated in the ShARe/CLEF eHealth Evaluation Lab Task 1. The goal of this task is to individuate mentions of disorders in free-text electronic health records and map disorders to SNOMED CT concepts in the UMLS metathesaurus. This paper details our participation to this ShARe/CLEF task. Our approaches are based on using the clinical natural language processing tool Metamap and Conditional Random Fields (CRF) to individuate mentions of disorders and then to map those to SNOMED CT concepts. Empirical results obtained on the 2013 ShARe/CLEF task highlight that our instance of Metamap (after ltering irrelevant semantic types), although achieving a high level of precision, is only able to identify a small amount of disorders (about 21% to 28%) from free-text health records. On the other hand, the addition of the CRF models allows for a much higher recall (57% to 79%) of disorders from free-text, without sensible detriment in precision. When evaluating the accuracy of the mapping of disorders to SNOMED CT concepts in the UMLS, we observe that the mapping obtained by our ltered instance of Metamap delivers state-of-the-art e ectiveness if only spans individuated by our system are considered (`relaxed' accuracy).

A cross-discipline and collaborative approach to identifying the predictors of environmentally friendly and health protective behaviour

Background. As a society, our interaction with the environment is having a negative impact on human health. For example, an increase in car use for short trips, over walking or cycling, has contributed to an increase in obesity, diabetes and poor heart health and also contributes to pollution, which is associated with asthma and other respiratory diseases. In order to change the nature of that interaction, to be more positive and healthy, it is recommended that individuals adopt a range of environmentally friendly behaviours (such as walking for transport and reducing the use of plastics). Effective interventions aimed at increasing such behaviours will need to be evidence based and there is a need for the rapid communication of information from the point of research, into policy and practice. Further, a number of health disciplines, including psychology and public health, share a common mission to promote health and well-being. Therefore, the objective of this project is to take a cross-discipline and collaborative approach to reveal psychological mechanisms driving environmentally friendly behaviour. This objective is further divided into three broad aims, the first of which is to take a cross-discipline and collaborative approach to research. The second aim is to explore and identify the salient beliefs which most strongly predict environmentally friendly behaviour. The third aim is to build an augmented model to explain environmentally friendly behaviour. The thesis builds on the understanding that an interdisciplinary collaborative approach will facilitate the rapid transfer of knowledge to inform behaviour change interventions. Methods. The application of this approach involved two surveys which explored the psycho-social predictors of environmentally friendly behaviour. Following a qualitative pilot study, and in collaboration with an expert panel comprising academics, industry professionals and government representatives, a self-administered, Theory of Planned Behaviour (TPB) based, mail survey was distributed to a random sample of 3000 residents of Brisbane and Moreton Bay Region (Queensland, Australia). This survey explored specific beliefs including attitudes, norms, perceived control, intention and behaviour, as well as environmental altruism and green identity, in relation to walking for transport and switching off lights when not in use. Following analysis of the mail survey data and based on feedback from participants and key stakeholders, an internet survey was employed (N=451) to explore two additional behaviours, switching off appliances at the wall when not in use, and shopping with reusable bags. This work is presented as a series of interrelated publications which address each of the research aims. Presentation of Findings. Chapter five of this thesis consists of a published paper which addresses the first aim of the research and outlines the collaborative and multidisciplinary approach employed in the mail survey. The paper argued that forging alliances with those who are in a position to immediately utilise the findings of research has the potential to improve the quality and timely communication of research. Illustrating this timely communication, Chapter six comprises a report presented to Moreton Bay Regional Council (MBRC). This report addresses aim's one and two. The report contains a summary of participation in a range of environmentally friendly behaviours and identifies the beliefs which most strongly predicted walking for transport and switching off lights (from the mail survey). These salient beliefs were then recommended as targets for interventions and included: participants believing that they might save money; that their neighbours also switch off lights; that it would be inconvenient to walk for transport and that their closest friend also walks for transport. Chapter seven also addresses the second aim and presents a published conference paper in which the salient beliefs predicting the four specified behaviours (from both surveys) are identified and potential applications for intervention are discussed. Again, a range of TPB based beliefs, including descriptive normative beliefs, were predictive of environmentally friendly behaviour. This paper was also provided to MBRC, along with recommendations for applying the findings. For example, as descriptive normative beliefs were consistently correlated with environmentally friendly behaviour, local councils could engage in marketing and interventions (workshops, letter box drops, internet promotions) which encourage parents and friends to model, rather than simply encourage, environmentally friendly behaviour. The final two papers, presented in Chapters eight and nine, addresses the third aim of the project. These papers each present two behaviours together to inform a TPB based theoretical model with which to predict environmentally friendly behaviour. A generalised model is presented, which is found to predict the four specific behaviours under investigation. The role of demographics was explored across each of the behaviour specific models. It was found that some behaviour's differ by age, gender, income or education. In particular, adjusted models predicted more of the variance in walking for transport amongst younger participants and females. Adjusted models predicted more variance in switching off lights amongst those with a bachelor degree or higher and predicted more variance in switching off appliances amongst those on a higher income. Adjusted models predicted more variance in shopping with reusable bags for males, people 40 years or older, those on a higher income and those with a bachelor degree or higher. However, model structure and general predictability was relatively consistent overall. The models provide a general theoretical framework from which to better understand the motives and predictors of environmentally friendly behaviour. Conclusion. This research has provided an example of the benefits of a collaborative interdisciplinary approach. It has identified a number of salient beliefs which can be targeted for social marketing campaigns and educational initiatives; and these findings, along with recommendations, have been passed on to a local council to be used as part of their ongoing community engagement programs. Finally, the research has informed a practical model, as well as behaviour specific models, for predicting sustainable living behaviours. Such models can highlight important core constructs from which targeted interventions can be designed. Therefore, this research represents an important step in undertaking collaborative approaches to improving population health through human-environment interactions.

Survival models of community tenure and length of hospital stay for the seriously mentally ill : a 10-year perspective

Objective: To examine the effects of personal and community characteristics, specifically race and rurality, on lengths of state psychiatric hospital and community stays using maximum likelihood survival analysis with a special emphasis on change over a ten year period of time. Data Sources: We used the administrative data of the Virginia Department of Mental Health, Mental Retardation, and Substance Abuse Services (DMHMRSAS) from 1982-1991 and the Area Resources File (ARF). Given these two sources, we constructed a history file for each individual who entered the state psychiatric system over the ten year period. Histories included demographic, treatment, and community characteristics. Study Design: We used a longitudinal, population-based design with maximum likelihood estimation of survival models. We presented a random effects model with unobserved heterogeneity that was independent of observed covariates. The key dependent variables were lengths of inpatient stay and subsequent length of community stay. Explanatory variables measured personal, diagnostic, and community characteristics, as well as controls for calendar time. Data Collection: This study used secondary, administrative, and health planning data. Principal Findings: African-American clients leave the community more quickly than whites. After controlling for other characteristics, however, race does not affect hospital length of stay. Rurality does not affect length of community stays once other personal and community characteristics are controlled for. However, people from rural areas have longer hospital stays even after controlling for personal and community characteristics. The effects of time are significantly smaller than expected. Diagnostic composition effects and a decrease in the rate of first inpatient admissions explain part of this reduced impact of time. We also find strong evidence for the existence of unobserved heterogeneity in both types of stays and adjust for this in our final models. Conclusions: Our results show that information on client characteristics available from inpatient stay records is useful in predicting not only the length of inpatient stay but also the length of the subsequent community stay. This information can be used to target increased discharge planning for those at risk of more rapid readmission to inpatient care. Correlation across observed and unobserved factors affecting length of stay has significant effects on the measurement of relationships between individual factors and lengths of stay. Thus, it is important to control for both observed and unobserved factors in estimation.

Bayesian models for spatio-temporal assessment of disease

This thesis has contributed to the advancement of knowledge in disease modelling by addressing interesting and crucial issues relevant to modelling health data over space and time. The research has led to the increased understanding of spatial scales, temporal scales, and spatial smoothing for modelling diseases, in terms of their methodology and applications. This research is of particular significance to researchers seeking to employ statistical modelling techniques over space and time in various disciplines. A broad class of statistical models are employed to assess what impact of spatial and temporal scales have on simulated and real data.

Bayesian hierarchical models for analysing spatial point-based data at a grid level: A comparison of approaches

Spatial data are now prevalent in a wide range of fields including environmental and health science. This has led to the development of a range of approaches for analysing patterns in these data. In this paper, we compare several Bayesian hierarchical models for analysing point-based data based on the discretization of the study region, resulting in grid-based spatial data. The approaches considered include two parametric models and a semiparametric model. We highlight the methodology and computation for each approach. Two simulation studies are undertaken to compare the performance of these models for various structures of simulated point-based data which resemble environmental data. A case study of a real dataset is also conducted to demonstrate a practical application of the modelling approaches. Goodness-of-fit statistics are computed to compare estimates of the intensity functions. The deviance information criterion is also considered as an alternative model evaluation criterion. The results suggest that the adaptive Gaussian Markov random field model performs well for highly sparse point-based data where there are large variations or clustering across the space; whereas the discretized log Gaussian Cox process produces good fit in dense and clustered point-based data. One should generally consider the nature and structure of the point-based data in order to choose the appropriate method in modelling a discretized spatial point-based data.

Understanding public response to a congestion charge : a random-effects ordered logit approach

Public acceptance is consistently listed as having an enormous impact on the implementation and success of a congestion charge scheme. This paper investigates public acceptance of such a scheme in Australia. Surveys were conducted in Brisbane and Melbourne, the two fastest growing Australian cities. Using an ordered logit modeling approach, the survey data including stated preferences were analyzed to pinpoint the important factors influencing people’s attitudes to a congestion charge and, in turn, to their transport mode choices. To accommodate the nature of, and to account for the resulting heterogeneity of the panel data, random effects were considered in the models. As expected, this study found that the amount of the congestion charge and the financial benefits of implementing it have a significant influence on respondents’ support for the charge and on the likelihood of their taking a bus to city areas. However, respondents’ current primary transport mode for travelling to the city areas has a more pronounced impact. Meanwhile, respondents’ perceptions of the congestion charge’s role in protecting the environment by reducing vehicle emissions, and of the extent to which the charge would mean that they travelled less frequently to the city for shopping or entertainment, also have a significant impact on their level of support for its implementation. We also found and explained notable differences across two cities. Finally, findings from this study have been fully discussed in relation to the literature.

Factors influencing robustness and effectiveness of conditional random fields in active learning frameworks

Active learning approaches reduce the annotation cost required by traditional supervised approaches to reach the same effectiveness by actively selecting informative instances during the learning phase. However, effectiveness and robustness of the learnt models are influenced by a number of factors. In this paper we investigate the factors that affect the effectiveness, more specifically in terms of stability and robustness, of active learning models built using conditional random fields (CRFs) for information extraction applications. Stability, defined as a small variation of performance when small variation of the training data or a small variation of the parameters occur, is a major issue for machine learning models, but even more so in the active learning framework which aims to minimise the amount of training data required. The factors we investigate are a) the choice of incremental vs. standard active learning, b) the feature set used as a representation of the text (i.e., morphological features, syntactic features, or semantic features) and c) Gaussian prior variance as one of the important CRFs parameters. Our empirical findings show that incremental learning and the Gaussian prior variance lead to more stable and robust models across iterations. Our study also demonstrates that orthographical, morphological and contextual features as a group of basic features play an important role in learning effective models across all iterations.

Multilevel competing risk models to evaluate the risk of nosocomial infection

Introduction Risk factor analyses for nosocomial infections (NIs) are complex. First, due to competing events for NI, the association between risk factors of NI as measured using hazard rates may not coincide with the association using cumulative probability (risk). Second, patients from the same intensive care unit (ICU) who share the same environmental exposure are likely to be more similar with regard to risk factors predisposing to a NI than patients from different ICUs. We aimed to develop an analytical approach to account for both features and to use it to evaluate associations between patient- and ICU-level characteristics with both rates of NI and competing risks and with the cumulative probability of infection. Methods We considered a multicenter database of 159 intensive care units containing 109,216 admissions (813,739 admission-days) from the Spanish HELICS-ENVIN ICU network. We analyzed the data using two models: an etiologic model (rate based) and a predictive model (risk based). In both models, random effects (shared frailties) were introduced to assess heterogeneity. Death and discharge without NI are treated as competing events for NI. Results There was a large heterogeneity across ICUs in NI hazard rates, which remained after accounting for multilevel risk factors, meaning that there are remaining unobserved ICU-specific factors that influence NI occurrence. Heterogeneity across ICUs in terms of cumulative probability of NI was even more pronounced. Several risk factors had markedly different associations in the rate-based and risk-based models. For some, the associations differed in magnitude. For example, high Acute Physiology and Chronic Health Evaluation II (APACHE II) scores were associated with modest increases in the rate of nosocomial bacteremia, but large increases in the risk. Others differed in sign, for example respiratory vs cardiovascular diagnostic categories were associated with a reduced rate of nosocomial bacteremia, but an increased risk. Conclusions A combination of competing risks and multilevel models is required to understand direct and indirect risk factors for NI and distinguish patient-level from ICU-level factors.

Monoclonal antibody screening of a phage-displayed random peptide library reveals mimotopes of chemokine receptor CCR5: Implications for the tertiary structure of the receptor and for an n-terminal binding site for HIV-1 gp120

The chemokine receptor CCR5 contains seven transmembrane-spanning domains. It binds chemokines and acts as co-receptor for macrophage (m)-tropic (or R5) strains of HIV-1. Monoclonal antibodies (mAb) to CCR5, 3A9 and 5C7, were used for biopanning a nonapeptide cysteine (C)-constrained phage-displayed random peptide library to ascertain contact residues and define tertiary structures of possible epitopes on CCR5. Reactivity of antibodies with phagotopes was established by enzyme-linked immunosorbent assay (ELISA). mAb 3A9 identified a phagotope C-HASIYDFGS-C (3A9/1), and 5C7 most frequently identified C-PHWLRDLRV-C (5C7/1). Corresponding peptides were synthesized. Phagotopes and synthetic peptides reacted in ELISA with corresponding antibodies and synthetic peptides inhibited antibody binding to the phagotopes. Reactivity by immunofluorescence of 3A9 with CCR5 was strongly inhibited by the corresponding peptide. Both mAb 3A9 and 5C7 reacted similarly with phagotopes and the corresponding peptide selected by the alternative mAb. The sequences of peptide inserts of phagotopes could be aligned as mimotopes of the sequence of CCR5. For phage 3A9/1, the motif SIYD aligned to residues at the N terminus and FG to residues on the first extracellular loop; for 5C7/1, residues at the N terminus, first extracellular loop, and possibly the third extracellular loop could be aligned and so would contribute to the mimotope. The synthetic peptides corresponding to the isolated phagotopes showed a CD4-dependent reactivity with gp120 of a primary, m-tropic HIV-1 isolate. Thus reactivity of antibodies raised to CCR5 against phage-displayed peptides defined mimotopes that reflect binding sites for these antibodies and reveal a part of the gp120 binding sites on CCR5.

Phagotopes derived by antibody screening of phage-displayed random peptide libraries vary in immunoreactivity: Studies using an exemplary monoclonal antibody, CII-C1, to type II collagen

Antibody screening of phage-displayed random peptide libraries to identify mimotopes of conformational epitopes is promising. However, because interpretations can be difficult, an exemplary system has been used in the present study to investigate whether variation in the peptide sequences of selected phagotopes corresponded with variation in immunoreactivity. The phagotopes, derived using a well-characterized monoclonal antibody, CII-C1, to a known conformational epitope on type II collagen, C1, were tested by direct and inhibition ELISA for reactivity with CII-C1. A multiple sequence alignment algorithm, PILEUP, was used to sort the peptides expressed by the phagotopes into clusters. A model was prepared of the C1 epitope on type II collagen. The 12 selected phagotopes reacted with CII-C1 by both direct ELISA (titres from < 100-11 200) and inhibition ELISA (20-100% inhibition); the reactivity varied according to the peptide sequence and assay format. The differences in reactivity between the phagotopes were mostly in accord with the alignment, by PILEUP, of the peptide sequences. The finding that the phagotopes functionally mimicked the C1 epitope on collagen was validated in that amino acids RRL at the amino terminal of many of the peptides were topographically demonstrable on the model of the C1 epitope. Notably, one phagotope that expressed the widely divergent peptide C-IAPKRHNSA-C also mimicked the C1 epitope, as judged by reactivity in each of the assays used: these included cross-inhibition of CII-C1 reactivity with each of the other phagotopes and inhibition by a synthetic peptide corresponding to that expressed by the most frequently selected phagotope, RRLPFGSQM. Thus, it has been demonstrated that multiple phage-displayed peptides can mimic the same epitope and that observed immunoreactivity of selected phagotopes with the selecting mAb can depend on the primary sequence of the expressed peptide and also on the assay format used.

Deeper and wider fully convolutional network coupled with conditional random fields for scene labeling

Deep convolutional neural networks (DCNNs) have been employed in many computer vision tasks with great success due to their robustness in feature learning. One of the advantages of DCNNs is their representation robustness to object locations, which is useful for object recognition tasks. However, this also discards spatial information, which is useful when dealing with topological information of the image (e.g. scene labeling, face recognition). In this paper, we propose a deeper and wider network architecture to tackle the scene labeling task. The depth is achieved by incorporating predictions from multiple early layers of the DCNN. The width is achieved by combining multiple outputs of the network. We then further refine the parsing task by adopting graphical models (GMs) as a post-processing step to incorporate spatial and contextual information into the network. The new strategy for a deeper, wider convolutional network coupled with graphical models has shown promising results on the PASCAL-Context dataset.