published_or_final_version / Civil Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / Statistics and Actuarial Science / Doctoral / Doctor of Philosophy

Spatial autocorrelation is commonly found in the Hedonic Pricing model for real estate prices, but little attention has been paid to identify the causes behind. The primary objective of this research is to examine the causes of spatial autocorrelation in housing prices. Observed autocorrelation is often attributable to the omission of important location characteristics in the modelling process. Since it is practically impossible to exhaustively include all location characteristics, some variables may eventually be omitted, leaving spatially autocorrelated residuals in the Hedonic Pricing model. This thesis proposes a new source of spatial autocorrelation: real estate market liquidity. We hypothesize that liquidity affects the geographical boundary within which buyers and sellers search for price information. When the “immediate vicinity” of a property has few transactions, buyers and sellers may have to search for price information from more distant locations. Therefore, low liquidity in the vicinity of a property should strengthen the spatial autocorrelation of real estate prices. A Spatial - Liquidity Hedonic Pricing (SLHP) model is proposed to test the above hypothesis. The SLHP model generalizes traditional spatial autoregressive models by making the spatial process liquidity dependent. When applied to the apartment market in Hong Kong, the model is operationalized by defining “immediate vicinity” as the building where the subject unit locates. Furthermore, the SLHP model recognizes that past transactions may affect current transactions, but not vice versa, so the spatial weight matrix is simply lower triangular. Under this condition, we have shown that the Maximum Likelihood Estimation is equivalent to the Ordinary Least Squares Estimation. This greatly simplifies the estimation procedures and reduces the empirical analysis to a feasible scale. Based on 15 500 transactions of residential units in Taikooshing, Hong Kong from 1992 to 2006, we conclude that while positive spatial autocorrelation is present in housing prices, its magnitude decreases when liquidity, as measured by the past transaction volume in the immediate vicinity of a subject unit, is high. In addition, we found that current prices are spatially correlated with transactions occurred up to the last three months only, reflecting the relatively high information efficiency of Hong Kong’s residential market. All these results are generally robust across a variety of distance, liquidity, and time weight specifications. This study establishes liquidity as a determinant of spatial autocorrelation in real estate prices. This is a new finding contributing to the economic literature on liquidity effects and technical literature on spatial estimation. Our results not only reveal the spatially dependent price formation process in the real estate market, but also have practical applications on the hedonic modelling of real estate prices for mass valuation and index construction. / published_or_final_version / Real Estate and Construction / Doctoral / Doctor of Philosophy

The demand for emergency ambulance services in Hong Kong is on the rise. Issues such as climate change, ageing population, constrained space, and limited resource capacity mean that the present way of meeting service demand by injecting more resources will reach its limit in the near future and unlikely to be sustainable. There is an urgent need to develop a more realistic forecast model to account for the anticipated demand for emergency ambulance services to enable better strategic planning of resources and more effective logistic arrangement. In this connection, the research objectives of this thesis include the following: 1. To examine relationships between weather and ambulance demand, with specific reference to temperature effects on demographic and admission characteristics of patients. 2. To establish a quantitative model for short-term (1-7 days ahead) forecast of ambulance demand in Hong Kong. 3. To estimate the longer-term demand for ambulance services by sub areas in Hong Kong, taking into account projected weather and population changes in 2019 and 2036. The research concurs with the findings of other researchers that temperature was the most important weather factor affecting the daily ambulance demand in 2006-2009, accounting for 49% of the demand variance. An even higher demand variance of 74% could be explained among people aged 65 and above. The incorporation of 1-7 day forecast data of the average temperature improved the forecast accuracy of daily ambulance demand on average by 33% in terms of R2 and 11% in terms of root mean square error (RMSE). Moreover, the forecast accuracy could be further improved by as much as 4% for both R2 and RMSE through spatial sub models. For demand projection of a longer-term, significant underestimation was observed if changes in the population demographics were not considered. The underestimation of annual ambulance demand for 2019 and 2036 was 16% and 38% respectively. The research has practical and methodological implications. First, the quantitative model for short-term forecast can inform demand in the next few days to enable logistic deployment of ambulance services beforehand, which, in turn, ensures that potential victims can be served in a swift and efficient manner. Second, the longer-term projection on the demand for ambulance services enables better preparation and planning for the expected rise in demand in time and space. Unbudgeted or unnecessary purchases of ambulances can be prevented without compromising preparedness and service quality. Third, the methodology is adaptable and the model can be reconstituted when more accurate projections on weather and population changes become available. / published_or_final_version / Geography / Doctoral / Doctor of Philosophy

The escalation with overdose control (EWOC) method is a popular modelbased dose finding design for phase I clinical trials. Dose finding for combined drugs has grown rapidly in oncology drug development. A two-dimensional EWOC design is proposed for dose finding with two agents in combination based on a four-parameter logistic regression model. During trial conduct, the posterior distribution of the maximum tolerated dose (MTD) combination is updated continuously in order to find the appropriate dose combination for each cohort of patients. The probability that the next dose combination exceeds the MTD combination can be controlled by a feasibility bound, which is based on a prespecified quantile for the MTD distribution such as to reduce the possibility of over-dosing. Dose escalation, de-escalation or staying at the same doses is determined by searching the MTD combination along rows and columns in a two-drug combination matrix. Simulation studies are conducted to examine the performance of the design under various practical scenarios, and illustrate it with a trial example. / published_or_final_version / Statistics and Actuarial Science / Master / Master of Philosophy

abstract / Public Health / Master / Master of Philosophy

Not available / text

Not available / text

(Uncorrected OCR) Abstract Abstract of thesis entitled Robust Statistics Based Adaptive Filtering Algorithms For Impulsive Noise Suppression Submitted by Yuexian Zou for the degree of Doctor of Philosophy at The University of Hong Kong in May 2000 The behavior of an adaptive filter is inherently decided by how its estimation error and the cost function are formulated under certain assumption of the involving signal statistics. This dissertation is concerned with the development of robust adaptive filtering in an impulsive noise environment based on the linear transversal filter (LTF) and the lattice-ladder filer (LLF) structures. Combining the linear adaptive filtering theory and robust statistics estimation techniques, two new cost functions, called the mean M -estimate error (MME) and the sum of weighted M -estimate error (SWME), are proposed. They can be taken as the generalizations of the well-known mean squared error (MSE) and the sum of weighted squares error (SWSE) cost functions when the involving signals are Gaussian. Based on the SWME cost function, the resulting optimal weight vector is governed by an M-estimate normal equation and a recursive least M -estimate (RLM) algorithm is derived. The RLM algorithm preserves the fast initial convergence, lower steady-state 11 Abstract derived. The RLM algorithm preserves the fast initial convergence, lower steady-state error and the robustness to the sudden system change of the recursive least squares (RLS) algorithm under Gaussian noise alone. Meanwhile, it has the ability to suppress impulse noise both in the desired and input signals. In addition, using the MME cost function, stochastic gradient based adaptive algorithms, named the least mean Mestimate (LMM) and its transform dOlnain version, the transform domain least mean Mestimate (TLMM) algorithms have been developed. The LMM and TLMM algorithms can be taken as the generalizations of the least-mean square (LMS) and transform domain normalized LMS (TLMS) algorithms, respectively. These two robust algorithms give similar performance as the LMS and TLMS algorithms under Gaussian noise alone and are able to suppress impulse noise appearing in the desired and input signals. It is noted that the performance and the computational complexity of the RLM, LMM and TLMM algorithms have a close relationship with the estimate of the threshold parameters for the M-estimate functions. A robust and effective recursive method has been suggested in this dissertation to estimate the variance of the estimation error and the required threshold parameters with certain confidence to suppress the impulsive noise. The mean and mean square convergence performances of the RLM and the LMM algorithms are evaluated, respectively, when the impulse noise is assumed to be contaminated Gaussian distribution. Motivated by the desirable features of the lattice-ladder filter, a new robust adaptive gradient lattice-ladder filtering algorithm is developed by minimizing an MME cost function together with an embedded robust impulse suppressing process, especially for impulses appearing in the filter input. The resultant robust gradient lattice-robust 111 Abstract normalized LMS (RGAL-RNLMS) algorithm perfonns comparably to the conventional GAL-NLMS algorithm under Gaussian noise alone; meanwhile, it has the capability of suppressing the adverse effects due to impulses in the input and the desired signals. The additional computational complexity compared to the GAL-NLMS algorithm is of O(Nw log Nw) + O(NfI log N,J . Extensive computer simulation studies are undertaken to evaluate the performance of the RLM, LMM, TLMM and the RGAL-RNLMS algorithms under the additive noise with either a contaminated Gaussian distribution or the symmetric alpha-stable (SaS ) distributions. The results substantiate the analysis and demonstrate the effectiveness and robustness of the developed robust adaptive filtering algorithms in suppressing impulsive noise both in the input and the desired signals of the adaptive filter. In conclusion, the proposed approaches in this dissertation present an attempt for developing robust adaptive filtering algorithms in impulsive noise environments and can be viewed as an extension of the linear adaptive filter theory. They may become reasonable and effective tools to solve adaptive filtering problems in a non-Gaussian environment in practice. IV / abstract / toc / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

(Uncorrected OCR) Abstract of the thesis entitled ON INSURANCE RISK MODELS WITH CORRELATED CLASSES OF BUSINESS submitted by Wu Xueyuan for the degree of Doctor of Philosophy at The University of Hong Kong in February 2004 In this thesis, we focus on ruin analysis of risk models wIth correlated classes of insurance business. Specifically, five risk models with different dependence relations between classes are introduced. For these models, various problems related to ruin probability are considered. vVe first study a continuous-time correlated aggregate clmms model with Poisson and Erlang risk processes. In this model, we assume that two classes of business are correlated through a common Erlang component in thelf claim-number processes. We derive an explicit expression for the mfimte-time survival probability of the assumed model when claim SIzes are exponentially distributed. For general claim-size distributions, we obtain some result for the infinite-time ruin probabIlIty, and present a numerical method for evaluating the probability of rum. Based on the continuous-tIme model of Yuen and "Vang (2002) with thin- ning correlatIOn, we propose a new dependence relatIOn with interaction between classes of business in the discrete-time case. Two dIscrete-time risk models with such a relation of dependence are studied. For the first interaction model: we investIgate the statIstical properties of the aggregate claIms for a family of claimnumber distributions. \Ve also compare the model with other existing models with correlated aggregate claIms in terms of the finite-time and infimte-time ruin probabllitles. The second model extends the interaction dependence to the case of the compound binomlal model with delayed claims. For this model, we develop a recursive method to compute the finite-time survival probabilities: and derive an explicit expression for the infinite-time survival probability in a special case. The last two risk models proposed in this thesis are the bivariate compound binomial model and the bivariate compound Poisson model. In the bivariate case: vanous definitions of ruin can be considered. For the bivariate compound binomial model, recursive algorithms for calculating several kinds of finite-time survival probability are presented and numerical examples are given. As for the bivariate compound Poisson model, we study the probabllity that at least one of the two classes of business will get ruined. Since this bivanate ruin probability is very dlfficult to deal with, we use the result of the bivariate compound binomial model to approximate the desired bivanate finite-time survlval probability. \Ve also obtain an upper bound for the infinite-time ruin probability via some association properties of the model. For a simplified version of the model, we examine 'l'l the mfimte-time ruin probability when claIm sizes are exponentially distributed. / abstract / toc / Statistics and Actuarial Science / Doctoral / Doctor of Philosophy