BURA Collection:http://bura.brunel.ac.uk/handle/2438/2352014-12-18T14:29:22Z2014-12-18T14:29:22ZNew regression methods for measures of central tendencyAristodemou, Katerinahttp://bura.brunel.ac.uk/handle/2438/92682014-12-03T03:00:36Z2014-01-01T00:00:00ZTitle: New regression methods for measures of central tendency
Authors: Aristodemou, Katerina
Abstract: Measures of central tendency have been widely used for summarising statistical data, with the mean being the most popular summary statistic. However, in reallife applications it is not always the most representative measure of central location, especially when dealing with data which is skewed or contains outliers. Alternative
statistics with less bias are the median and the mode. Median and quantile regression has been used in different fields to examine the effect of factors at different points of the distribution. Mode estimation, on the other hand, has found many applications in cases where the analysis focuses on obtaining information about the most typical value or pattern. This thesis demonstrates that mode also plays an important role in the analysis of big data, which is becoming increasingly important in many sectors of the global economy.
However, mode regression has not been widely applied, even though there is a clear conceptual benefit, due to the computational and theoretical limitations of the existing estimators. Similarly, despite the popularity of the binary quantile regression model, computational straight forward estimation techniques do not exist.
Driven by the demand for simple, well-found and easy to implement inference tools, this thesis develops a series of new regression methods for mode and binary quantile regression. Chapter 2 deals with mode regression methods from the Bayesian perspective and presents one parametric and two non-parametric methods of inference. Chapter 3 demonstrates a mode-based, fast pattern-identification method for big data and proposes the first fully parametric mode regression method, which effectively uncovers the dependency of typical patterns on a number of covariates. The proposed approach is demonstrated through the analysis of a decade-long dataset on the Body Mass Index and associated factors, taken from the Health Survey for England. Finally, Chapter 4 presents an alternative binary quantile regression approach, based on the nonlinear least asymmetric weighted squares, which can be implemented using standard statistical packages and guarantees a unique solution.
Description: This thesis was submitted for the degree of Doctor of Philosophy and awarded by Brunel University2014-01-01T00:00:00ZStructure and dynamics of evolving complex networksColman, Ewanhttp://bura.brunel.ac.uk/handle/2438/92082014-11-13T13:04:47Z2014-01-01T00:00:00ZTitle: Structure and dynamics of evolving complex networks
Authors: Colman, Ewan
Abstract: The analysis of large disordered complex networks has recently received enormous attention motivated by both academic and commercial interest. The most important results in this discipline have come from the analysis of stochastic models which mimic the growth and evolution of real networks as they change over time. The purpose of this thesis is to introduce various novel processes which dictate the development of a network on a small scale, and use techniques learned from statistical physics to derive the dynamical and structural properties of the network on the macroscopic scale. We introduce each model as a set of mechanisms determining how a network changes over a small period in time, from these rules we derive several topological
properties of the network after many iterations, most notably the degree distribution. 1. In the rst mechanism, nodes are introduced and linked to older nodes in the network in such a way as to create triangles and maintain a high level of clustering. The mechanism resembles the growth of a citation network and we demonstrate analytically that the mechanism introduced su ces to explain the power-law form commonly found in citation distributions. 2. The second mechanism involves edge rewiring processes - detaching one end of an edge and reattaching it, either to a random node anywhere in the network or to one selected locally. 3. We analyse a variety of processes based around a novel fragmentation mechanism. 4. The nal model concerns the problem of nding the electrical resistance across a network. The network grows as a random tree, as it grows the distribution of resistance converges towards a steady state solution. We nd an application of the relatively recent concept of a random Fibonacci sequence in deriving the rate of convergence of the mean.
Description: This thesis was submitted for the degree of Doctor of Philosophy and awarded by Brunel University2014-01-01T00:00:00ZRegularized and robust regression methods for high dimensional dataHashem, Hussein Abdulahmanhttp://bura.brunel.ac.uk/handle/2438/91972014-11-01T11:38:39Z2014-01-01T00:00:00ZTitle: Regularized and robust regression methods for high dimensional data
Authors: Hashem, Hussein Abdulahman
Abstract: Recently, variable selection in high-dimensional data has attracted much research interest. Classical stepwise subset selection methods are widely used in practice, but when the number of predictors is large these methods are difficult to implement. In these cases, modern regularization methods have become a popular choice as they perform variable selection and parameter estimation simultaneously. However, the estimation procedure becomes more difficult and challenging when the data suffer from outliers or when the assumption of normality is violated such as in the case of heavy-tailed errors. In these cases, quantile regression is the most appropriate method to use. In this thesis we combine these two classical approaches together to produce regularized quantile regression methods. Chapter 2 shows a comparative simulation study of regularized and robust regression methods when the response variable is continuous. In chapter 3, we develop a quantile regression model with a group lasso penalty for binary response data when the predictors have a grouped structure and when the data suffer from outliers. In chapter 4, we extend this method to the case of censored response variables. Numerical examples on simulated and real data are used to evaluate the performance of the proposed methods in comparisons with other existing methods.
Description: This thesis was submitted for the degree of Doctor of Philosophy and awarded by Brunel University London.2014-01-01T00:00:00ZA behavioural approach to financial portfolio selection problem: an empirical study using heuristicsGrishina, Ninahttp://bura.brunel.ac.uk/handle/2438/91732014-11-01T11:38:58Z2014-01-01T00:00:00ZTitle: A behavioural approach to financial portfolio selection problem: an empirical study using heuristics
Authors: Grishina, Nina
Abstract: The behaviourally based portfolio selection problem with investor's loss aversion and risk aversion biases in portfolio choice under uncertainty are studied. The main results of this work are developed heuristic approaches for the prospect theory and cumulative prospect theory models proposed by Kahneman and Tversky in 1979 and 1992 as well as an empirical comparative analysis of these models and the traditional mean variance and index tracking models. The crucial assumption is that behavioural features of the (cumulative) prospect theory model provide better downside protection than traditional approaches to the portfolio selection problem. In this research the large scale computational results for the (cumulative) prospect theory model have been obtained. Previously, as far as we aware, only small laboratory (2-3 arti cial assets) tests has been presented in the literature. In order to investigate empirically the performance of the behaviourally based models, a differential evolution algorithm and a genetic algorithm which are capable to
deal with large universe of assets have been developed. The speci c breeding and mutation as well as normalisation have been implemented in the algorithms. A tabulated comparative analysis of the algorithms' parameter choice is presented. The performance of the studied models have been tested out-of-sample in different conditions using the bootstrap method as well as simulation of the distribution of a growing market and simulation of the t-distribution with fat tails which characterises the dynamics of a decreasing or crisis market. A cardinality and CVaR constraints have been implemented to the basic mean variance and prospect theory models. The comparative analysis of the empirical results has been made using several criteria such as CPU time, ratio between mean portfolio return and
standart deviation, mean portfolio return, standard deviation , VaR and CVaR as alternative measures of risk. The strong in
uence of the reference point, loss aversion and risk aversion on the prospect theory model's results have been found. The prospect theory model with the reference point being the index is compared to the index tracking model. The portfolio diversi cation bene t has been found. However, the aggressive behaviour in terms of returns of the prospect theory model with the reference point being the index leads to worse performance of this model in a bearish market compared to the index tracking model. The tabulated comparative analysis of the performance of all studied models is provided in this research for in-sample and out-of-sample tests.
Description: This thesis was submitted for the degree of Doctor of Philosophy and awarded by Brunel University2014-01-01T00:00:00Z