Using Bayesian methods for the estimation of uncertainty in complex statistical models

The research objectives of this thesis were to contribute to Bayesian statistical methodology by contributing to risk assessment statistical methodology, and to spatial and spatio-temporal methodology, by modelling error structures using complex hierarchical models. Specifically, I hoped to consider two applied areas, and use these applications as a springboard for developing new statistical methods as well as undertaking analyses which might give answers to particular applied questions. Thus, this thesis considers a series of models, firstly in the context of risk assessments for recycled water, and secondly in the context of water usage by crops. The research objective was to model error structures using hierarchical models in two problems, namely risk assessment analyses for wastewater, and secondly, in a four dimensional dataset, assessing differences between cropping systems over time and over three spatial dimensions. The aim was to use the simplicity and insight afforded by Bayesian networks to develop appropriate models for risk scenarios, and again to use Bayesian hierarchical models to explore the necessarily complex modelling of four dimensional agricultural data. The specific objectives of the research were to develop a method for the calculation of credible intervals for the point estimates of Bayesian networks; to develop a model structure to incorporate all the experimental uncertainty associated with various constants thereby allowing the calculation of more credible credible intervals for a risk assessment; to model a single day’s data from the agricultural dataset which satisfactorily captured the complexities of the data; to build a model for several days’ data, in order to consider how the full data might be modelled; and finally to build a model for the full four dimensional dataset and to consider the timevarying nature of the contrast of interest, having satisfactorily accounted for possible spatial and temporal autocorrelations. This work forms five papers, two of which have been published, with two submitted, and the final paper still in draft. The first two objectives were met by recasting the risk assessments as directed, acyclic graphs (DAGs). In the first case, we elicited uncertainty for the conditional probabilities needed by the Bayesian net, incorporated these into a corresponding DAG, and used Markov chain Monte Carlo (MCMC) to find credible intervals, for all the scenarios and outcomes of interest. In the second case, we incorporated the experimental data underlying the risk assessment constants into the DAG, and also treated some of that data as needing to be modelled as an ‘errors-invariables’ problem [Fuller, 1987]. This illustrated a simple method for the incorporation of experimental error into risk assessments. In considering one day of the three-dimensional agricultural data, it became clear that geostatistical models or conditional autoregressive (CAR) models over the three dimensions were not the best way to approach the data. Instead CAR models are used with neighbours only in the same depth layer. This gave flexibility to the model, allowing both the spatially structured and non-structured variances to differ at all depths. We call this model the CAR layered model. Given the experimental design, the fixed part of the model could have been modelled as a set of means by treatment and by depth, but doing so allows little insight into how the treatment effects vary with depth. Hence, a number of essentially non-parametric approaches were taken to see the effects of depth on treatment, with the model of choice incorporating an errors-in-variables approach for depth in addition to a non-parametric smooth. The statistical contribution here was the introduction of the CAR layered model, the applied contribution the analysis of moisture over depth and estimation of the contrast of interest together with its credible intervals. These models were fitted using WinBUGS [Lunn et al., 2000]. The work in the fifth paper deals with the fact that with large datasets, the use of WinBUGS becomes more problematic because of its highly correlated term by term updating. In this work, we introduce a Gibbs sampler with block updating for the CAR layered model. The Gibbs sampler was implemented by Chris Strickland using pyMCMC [Strickland, 2010]. This framework is then used to consider five days data, and we show that moisture in the soil for all the various treatments reaches levels particular to each treatment at a depth of 200 cm and thereafter stays constant, albeit with increasing variances with depth. In an analysis across three spatial dimensions and across time, there are many interactions of time and the spatial dimensions to be considered. Hence, we chose to use a daily model and to repeat the analysis at all time points, effectively creating an interaction model of time by the daily model. Such an approach allows great flexibility. However, this approach does not allow insight into the way in which the parameter of interest varies over time. Hence, a two-stage approach was also used, with estimates from the first-stage being analysed as a set of time series. We see this spatio-temporal interaction model as being a useful approach to data measured across three spatial dimensions and time, since it does not assume additivity of the random spatial or temporal effects.

Analysis for the use of cumulative plots for travel time estimation on signalized network

This paper provides fundamental understanding for the use of cumulative plots for travel time estimation on signalized urban networks. Analytical modeling is performed to generate cumulative plots based on the availability of data: a) Case-D, for detector data only; b) Case-DS, for detector data and signal timings; and c) Case-DSS, for detector data, signal timings and saturation flow rate. The empirical study and sensitivity analysis based on simulation experiments have observed the consistency in performance for Case-DS and Case-DSS, whereas, for Case-D the performance is inconsistent. Case-D is sensitive to detection interval and signal timings within the interval. When detection interval is integral multiple of signal cycle then it has low accuracy and low reliability. Whereas, for detection interval around 1.5 times signal cycle both accuracy and reliability are high.

Representational change and strategy use in children's number line estimation during the first years of primary school

Background: The objective of this study was to scrutinize number line estimation behaviors displayed by children in mathematics classrooms during the first three years of schooling. We extend existing research by not only mapping potential logarithmic-linear shifts but also provide a new perspective by studying in detail the estimation strategies of individual target digits within a number range familiar to children. Methods: Typically developing children (n = 67) from Years 1 – 3 completed a number-to-position numerical estimation task (0-20 number line). Estimation behaviors were first analyzed via logarithmic and linear regression modeling. Subsequently, using an analysis of variance we compared the estimation accuracy of each digit, thus identifying target digits that were estimated with the assistance of arithmetic strategy. Results: Our results further confirm a developmental logarithmic-linear shift when utilizing regression modeling; however, uniquely we have identified that children employ variable strategies when completing numerical estimation, with levels of strategy advancing with development. Conclusion: In terms of the existing cognitive research, this strategy factor highlights the limitations of any regression modeling approach, or alternatively, it could underpin the developmental time course of the logarithmic-linear shift. Future studies need to systematically investigate this relationship and also consider the implications for educational practice.

Estimation on disease burden related to hepatitis B virus infection in Shandong province of China [山东省乙型肝炎病毒感染所致相关疾病的疾病负担研究]

Objective: To comprehensively measure the burden of hepatitis B, liver cirrhosis and liver cancer in Shandong province, using disability-adjusted life years (DALYs) to estimate the disease burden attribute to hepatitis B virus (HBV)infection. Methods: Based on the mortality data of hepatitis B, liver cirrhosis and liver cancer derived from the third National Sampling Retrospective Survey for Causes of Death during 2004 and 2005, the incidence data of hepatitis B and the prevalence and the disability weights of liver cancer gained from the Shandong Cancer Prevalence Sampling Survey in 2007, we calculated the years of life lost (YLLs), years lived with disability (YLDs) and DALYs of three diseases following the procedures developed for the global burden of disease (GBD) study to ensure the comparability. Results: The total burden for hepatitis B, liver cirrhosis and liver cancer were 211 616 (39 377 YLLs and 172 239 YLDs), 16 783 (13 497 YLLs and 3286 YLDs) and 247 795 (240 236 YLLs and 7559 YLDs) DALYs in 2005 respectively, and men were 2.19, 2.36 and 3.16 times as that for women, respectively in Shandong province. The burden for hepatitis B was mainly because of disability (81.39%). However, most burden on liver cirrhosis and liver cancer were due to premature death (80.42% and 96.95%). The burden of each patient related to hepatitis B, liver cirrhosis and liver cancer were 4.8, 13.73 and 11.11 respectively. Conclusion: Hepatitis B, liver cirrhosis and liver cancer caused considerable burden to the people living in Shandong province, indicating that the control of hepatitis B virus infection would bring huge potential benefits.

Estimation on the mortality and disease burden attributed to selected risk factors in Shandong province [山东省主要危险因素的归因死亡和疾病负担分析]

Objective: To determine the major health related risk factors and provide evidence for policy-making,using health burden analysis on selected factors among general population from Shandong province. Methods: Based on data derived from the Third Death of Cause Sampling Survey in Shandong. Years of life lcrat(YLLs),yearS Iived with disability(YLDs)and disability-adjusted life years(DALYs) were calculated according to the GBD ethodology.Deaths and DALYs attributed to the selected risk factors were than estimated together with the PAF data from GBD 2001 study.The indirect method was employed to estimate the YLDs. Results: 51.09%of the total dearlls and 31.83%of the total DALYs from the Shandong population were resulted from the 19 selected risk factors.High blood pre.ure,smoking,low fruit and vegetable intake,aleohol consumption,indoor smoke from solid fuels,high cholesterol,urban air pollution, physical inactivity,overweight and obesity and unsafe injections in health care settings were identified as the top 10 risk faetors for mortality which together caused 50.21%of the total deaths.Alcohol use,smoking,high blood pressure,Low fruit and vegetable intake, indoor smoke from solid fuels, overweight and obesity,high cholesterol, physical inactivity,urban air pollution and iron-deficiency anemia were proved as the top 10 risk factors related to disease burden and were responsible for 29.04%of the total DALYs. Conclusion: Alcohol use.smoking and high blood pressure were determined as the major risk factors which influencing the health of residents in Shandong. The mortality and burden of disease could be reduced significantly if these major factors were effectively under control.

Estimation of energy saving of commercial building by living wall and green facade in sub-tropical climate of Australia

This paper investigates energy saving potential of commercial building by living wall and green façade system using Envelope Thermal Transfer Value (ETTV) equation in Sub-tropical climate of Australia. Energy saving of four commercial buildings was quantified by applying living wall and green façade system to the west facing wall. A field experimental facility, from which temperature data of living wall system was collected, was used to quantify wall temperatures and heat gain under controlled conditions. The experimental parameters were accumulated with extensive data of existing commercial building to quantify energy saving. Based on temperature data of living wall system comprised of Australian native plants, equivalent temperature of living wall system has been computed. Then, shading coefficient of plants in green façade system has been included in mathematical equation and in graphical analysis. To minimize the air-conditioned load of commercial building, therefore to minimize the heat gain of commercial building, an analysis of building heat gain reduction by living wall and green façade system has been performed. Overall, cooling energy performance of commercial building before and after living wall and green façade system application has been examined. The quantified energy saving showed that only living wall system on opaque part of west facing wall can save 8-13 % of cooling energy consumption where as only green façade system on opaque part of west facing wall can save 9.5-18% cooling energy consumption of commercial building. Again, green façade system on fenestration system on west facing wall can save 28-35 % of cooling energy consumption where as combination of both living wall on opaque part of west facing wall and green façade on fenestration system on west facing wall can save 35-40% cooling energy consumption of commercial building in sub-tropical climate of Australia.

Integrated approach for HP-LNG pump diagnosis and prognostics based on health state probability estimation

This paper presents an innovative prognostics model based on health state probability estimation embedded in the closed loop diagnostic and prognostic system. To employ an appropriate classifier for health state probability estimation in the proposed prognostic model, the comparative intelligent diagnostic tests were conducted using five different classifiers applied to the progressive fault levels of three faults in HP-LNG pump. Two sets of impeller-rubbing data were employed for the prediction of pump remnant life based on estimation of discrete health state probability using an outstanding capability of SVM and a feature selection technique. The results obtained were very encouraging and showed that the proposed prognosis system has the potential to be used as an estimation tool for machine remnant life prediction in real life industrial applications.

Change Point Estimation in Monitoring Survival Time

Precise identification of the time when a change in a hospital outcome has occurred enables clinical experts to search for a potential special cause more effectively. In this paper, we develop change point estimation methods for survival time of a clinical procedure in the presence of patient mix in a Bayesian framework. We apply Bayesian hierarchical models to formulate the change point where there exists a step change in the mean survival time of patients who underwent cardiac surgery. The data are right censored since the monitoring is conducted over a limited follow-up period. We capture the effect of risk factors prior to the surgery using a Weibull accelerated failure time regression model. Markov Chain Monte Carlo is used to obtain posterior distributions of the change point parameters including location and magnitude of changes and also corresponding probabilistic intervals and inferences. The performance of the Bayesian estimator is investigated through simulations and the result shows that precise estimates can be obtained when they are used in conjunction with the risk-adjusted survival time CUSUM control charts for different magnitude scenarios. The proposed estimator shows a better performance where a longer follow-up period, censoring time, is applied. In comparison with the alternative built-in CUSUM estimator, more accurate and precise estimates are obtained by the Bayesian estimator. These superiorities are enhanced when probability quantification, flexibility and generalizability of the Bayesian change point detection model are also considered.

Volume estimation from multiplanar 2D ultrasound images using a remote electromagnetic position and orientation sensor

A system is described for calculating volume from a sequence of multiplanar 2D ultrasound images. Ultrasound images are captured using a video digitising card (Hauppauge Win/TV card) installed in a personal computer, and regions of interest transformed into 3D space using position and orientation data obtained from an electromagnetic device (Polbemus, Fastrak). The accuracy of the system was assessed by scanning 10 water filled balloons (13-141 ml), 10 kidneys (147  200 ml) and 16 fetal livers (8  37 ml) in water using an Acuson 128XP/10 (5 MHz curvilinear probe). Volume was calculated using the ellipsoid, planimetry, tetrahedral and ray tracing methods and compared with the actual volume measured by weighing (balloons) and water displacement (kidneys and livers). The mean percentage error for the ray tracing method was 0.9 ± 2.4%, 2.7 ± 2.3%, 6.6 ± 5.4% for balloons, kidneys and livers, respectively. So far the system has been used clinically to scan fetal livers and lungs, neonate brain ventricles and adult prostate glands.

In vitro estimation of foetal liver volume using ultrasound, x-ray computed tomography and magnetic resonance imaging

Sixteen formalin-fixed foetal livers were scanned in vitro using a new system for estimating volume from a sequence of multiplanar 2D ultrasound images. Three different scan techniques were used (radial, parallel and slanted) and four volume estimation algorithms (ellipsoid, planimetry, tetrahedral and ray tracing). Actual liver volumes were measured by water displacement. Twelve of the sixteen livers also received x-ray computed tomography (CT) and magnetic resonance (MR) scans and the volumes were calculated using voxel counting and planimetry. The percentage accuracy (mean ± SD) was 5.3 ± 4.7%, −3.1 ± 9.6% and −0.03 ± 9.7% for ultrasound (radial scans, ray volumes), MR and CT (voxel counting) respectively. The new system may be useful for accurately estimating foetal liver volume in utero.

An estimation of crash risks using a log-linear model and the induced exposure technique

Prevention and safety promotion programmes. Traditionally, in-depth investigations of crash risks are conducted using exposure controlled study or case-control methodology. However, these studies need either observational data for control cases or exogenous exposure data like vehicle-kilometres travel, entry flow or product of conflicting flow for a particular traffic location, or a traffic site. These data are not readily available and often require extensive data collection effort on a system-wide basis. Aim: The objective of this research is to propose an alternative methodology to investigate crash risks of a road user group in different circumstances using readily available traffic police crash data. Methods: This study employs a combination of a log-linear model and the quasi-induced exposure technique to estimate crash risks of a road user group. While the log-linear model reveals the significant interactions and thus the prevalence of crashes of a road user group under various sets of traffic, environmental and roadway factors, the quasi-induced exposure technique estimates relative exposure of that road user in the same set of explanatory variables. Therefore, the combination of these two techniques provides relative measures of crash risks under various influences of roadway, environmental and traffic conditions. The proposed methodology has been illustrated using Brisbane motorcycle crash data of five years. Results: Interpretations of results on different combination of interactive factors show that the poor conspicuity of motorcycles is a predominant cause of motorcycle crashes. Inability of other drivers to correctly judge the speed and distance of an oncoming motorcyclist is also evident in right-of-way violation motorcycle crashes at intersections. Discussion and Conclusions: The combination of a log-linear model and the induced exposure technique is a promising methodology and can be applied to better estimate crash risks of other road users. This study also highlights the importance of considering interaction effects to better understand hazardous situations. A further study on the comparison between the proposed methodology and case-control method would be useful.

Accuracy of rate estimation using relaxed-clock models with a critical focus on the early metazoan radiation

In recent years, a number of phylogenetic methods have been developed for estimating molecular rates and divergence dates under models that relax the molecular clock constraint by allowing rate change throughout the tree. These methods are being used with increasing frequency, but there have been few studies into their accuracy. We tested the accuracy of several relaxed-clock methods (penalized likelihood and Bayesian inference using various models of rate change) using nucleotide sequences simulated on a nine-taxon tree. When the sequences evolved with a constant rate, the methods were able to infer rates accurately, but estimates were more precise when a molecular clock was assumed. When the sequences evolved under a model of autocorrelated rate change, rates were accurately estimated using penalized likelihood and by Bayesian inference using lognormal and exponential models of rate change, while other models did not perform as well. When the sequences evolved under a model of uncorrelated rate change, only Bayesian inference using an exponential rate model performed well. Collectively, the results provide a strong recommendation for using the exponential model of rate change if a conservative approach to divergence time estimation is required. A case study is presented in which we use a simulation-based approach to examine the hypothesis of elevated rates in the Cambrian period, and it is found that these high rate estimates might be an artifact of the rate estimation method. If this bias is present, then the ages of metazoan divergences would be systematically underestimated. The results of this study have implications for studies of molecular rates and divergence dates.

Accounting for calibration uncertainty in phylogenetic estimation of evolutionary divergence times

The estimation of phylogenetic divergence times from sequence data is an important component of many molecular evolutionary studies. There is now a general appreciation that the procedure of divergence dating is considerably more complex than that initially described in the 1960s by Zuckerkandl and Pauling (1962, 1965). In particular, there has been much critical attention toward the assumption of a global molecular clock, resulting in the development of increasingly sophisticated techniques for inferring divergence times from sequence data. In response to the documentation of widespread departures from clocklike behavior, a variety of local- and relaxed-clock methods have been proposed and implemented. Local-clock methods permit different molecular clocks in different parts of the phylogenetic tree, thereby retaining the advantages of the classical molecular clock while casting off the restrictive assumption of a single, global rate of substitution (Rambaut and Bromham 1998; Yoder and Yang 2000).

Branch-length estimation bias misleads molecular dating for a vertebrate mitochondrial phylogeny

Despite recent methodological advances in inferring the time-scale of biological evolution from molecular data, the fundamental question of whether our substitution models are sufficiently well specified to accurately estimate branch-lengths has received little attention. I examine this implicit assumption of all molecular dating methods, on a vertebrate mitochondrial protein-coding dataset. Comparison with analyses in which the data are RY-coded (AG → R; CT → Y) suggests that even rates-across-sites maximum likelihood greatly under-compensates for multiple substitutions among the standard (ACGT) NT-coded data, which has been subject to greater phylogenetic signal erosion. Accordingly, the fossil record indicates that branch-lengths inferred from the NT-coded data translate into divergence time overestimates when calibrated from deeper in the tree. Intriguingly, RY-coding led to the opposite result. The underlying NT and RY substitution model misspecifications likely relate respectively to “hidden” rate heterogeneity and changes in substitution processes across the tree, for which I provide simulated examples. Given the magnitude of the inferred molecular dating errors, branch-length estimation biases may partly explain current conflicts with some palaeontological dating estimates.