Thesis (Ph. D.)--Drexel University, 2007. / Includes abstract and vita. Includes bibliographical references (leaves 96-103).

Thesis (M.Phil.)--Hong Kong University of Science and Technology, 2007. / Includes bibliographical references (leaves 47-51). Also available in electronic version.

In this expository paper we study the stationary measures of a stochastic process called nearest neighbor random walk on Z, and further we describe conditions for these measures to have the stronger property of reversibility. We consider both the cases of symmetric and non-symmetric random walk.

An error in the Gaussian random variate generator by Kinderman and Ramage is described that results in the generation of random variates with an incorrect distribution. An additional statement that corrects the original algorithm is given. / Series: Preprint Series / Department of Applied Statistics and Data Processing

Depressed and suicidal patients have difficulty in recollecting specific autobiographical events. In response to cue words they tend to generate summarised or general memories instead of specific events. The objectives of this thesis are to explore the mechanisms underlying the production of specific and general autobiographical memories in a non clinical population. The roles of imagery and working memory in the generation of autobiographical memories were investigated. Four experiments examined how manipulating the imageability of the cue affected subsequent retrieval in autobiographical memory. The results show that cues high in imageability facilitated access to specific memories and that visual imageability was the most significant piedictor of memory specificity compared to a range of other perceptual modalities. The effect of an experimental manipulation on retrieval style was examined by instructing participants to retrieve specific events or general events using high or low imageable words to cue memories. The results show that induction. of a generic retrieval style reduced the specificity of images of future events. This models clinical findings with depressed and suicidal patients and suggests that associations between memory retrieval and future imaging share common intermediate pathways. A further experiment suggested that the image ability effects mediating the construction of specific memories may be in part due to the predicability of such retrieval cues. The hypothesis that retrieval of specific autobiographical memories is more effortful compared to the retrieval of general memories was also investigated using a dual task paradigm. Although central executive function has been implicated many times in the monitoring of autobiographical retrieval, no direct assessment of executive capacity during retrieval has been made. The results showed no significant difference in the randomness of a keypressing task when specific or general autobiographical memories were retrieved in response to either high or low imageable cue words. A direct retrieval hypothesis was proposed whereby cues directly accessed specific events in autobiographical memory and the adoption of such a strategy enabled participants to maintain performance on the secondary task.

In 1986, Boguslaw Tomaszewski asked the following question: Consider n real numbers a1, . . . , an such that the sum of their squares is 1. Of the 2n expressions |ε1a1 + · · · + εnan| with εi = ±1, can there be more with value > 1 than with value ≤ 1? Apart from being of intrinsic interest in probability, an answer to this conjecture would also have applications in quadratic programming. However, even after more than thirty years the conjecture is still unsolved. In this thesis we settle a special case of the conjecture - we prove that the conjecture holds for vectors of the form (α, δ, . . . , δ) of sufficiently large dimension. This generalizes earlier result which showed that the conjecture holds for vectors of the form (δ, . . . , δ). 1

abstract: Random Forests is a statistical learning method which has been proposed for propensity score estimation models that involve complex interactions, nonlinear relationships, or both of the covariates. In this dissertation I conducted a simulation study to examine the effects of three Random Forests model specifications in propensity score analysis. The results suggested that, depending on the nature of data, optimal specification of (1) decision rules to select the covariate and its split value in a Classification Tree, (2) the number of covariates randomly sampled for selection, and (3) methods of estimating Random Forests propensity scores could potentially produce an unbiased average treatment effect estimate after propensity scores weighting by the odds adjustment. Compared to the logistic regression estimation model using the true propensity score model, Random Forests had an additional advantage in producing unbiased estimated standard error and correct statistical inference of the average treatment effect. The relationship between the balance on the covariates' means and the bias of average treatment effect estimate was examined both within and between conditions of the simulation. Within conditions, across repeated samples there was no noticeable correlation between the covariates' mean differences and the magnitude of bias of average treatment effect estimate for the covariates that were imbalanced before adjustment. Between conditions, small mean differences of covariates after propensity score adjustment were not sensitive enough to identify the optimal Random Forests model specification for propensity score analysis. / Dissertation/Thesis / Ph.D. Psychology 2013

This thesis is about graph-indexed random walks, Lipschitz mappings and graph homo- morphisms. It discusses connections between these notions, surveys the existing results, and shows new results. Graph homomorphism is an adjacency-preserving mapping between two graphs. Our main objects of study are graph homomorphisms to an infinite path. We are interested in two parameters: maximum range and average range. The average range of a graph is the expected size of the image of a uniformly picked random homomorphism to an infinite path. We obtain formulas for several graph classes and investigate main conjectures on this parameter. For maximum range parameter we show a general formula and an algorithm to compute it for general graphs. Besides that, we study the problem of extending a prescribed partial graph homomorphism to a full graph homomorphism. We show that this problem is polynomial in some cases. 1

In this thesis we investigate the geometric and algorithmic aspects of the random-cluster model, a correlated bond percolation model of great importance in the field of mathematics and statistical mechanics. We focus on the computational and statistical efficiency of the single-bond or heat-bath Markov chain for the random-cluster model and develop algorithmic techniques that allow for an improvement from a previously known polynomial to a poly-logarithmic runtime scaling of updates for general graphs. The interplay between the (critical) cluster structure of the random-cluster model and algorithmic, as well as statistical, efficiencies is considered, leading to new exact identities. A complementary analysis of certain fragility properties of the Fortuin-Kasteleyn clusters provides new insights into fragmentation phenomena, culminating in a revised scaling relation for a related fragmentation power law exponent, previously only shown for the marginal bond percolation case. By utilising the established structural results, a dynamic fragmentation process is studied that allows for an extraction of characteristics of the equilibrium cluster structure by a careful analysis of the limiting fragments, as well as the entire evolution of the fragmentation process. Besides focussing on structural and computational aspects, in this dissertation we also analyse the efficiency of the coupling from the past perfect sampling algorithm for the random-cluster model via large-scale numerical simulations. Two key results are the particular, close to optimal, efficiency in the off-critical setting and the intriguing observation of its superiority compared to the alternative Chayes-Machta-Swendsen-Wang approach in three dimensions. Governed by a random runtime, the efficiency of the coupling from the past algorithm depends crucially on the fluctuations of the runtime. In this connection a compelling appearance of universal Gumbel fluctuations in the distribution of the runtime of the coupling from the past algorithm is established, both at and off criticality. Fluctuations at a tricritical point and at a discontinuous phase transition are shown to deviate from this Gumbel law. The above findings in two and three dimensions are supported by a rigorous analysis of certain aspects of the algorithm in one dimension, including a proof of the limiting Gumbel law.

D.Phil. (Mathematical Statistics) / Fractional Brownian motion and its increment process, fractional Gaussian noise, are syn- onymous with the concept of long range dependence. A strictly stationary time series is said to exhibit long range dependence or long memory if its autocorrelations decrease to zero as a power of the lag, but their sum over all lags is not absolutely convergent. This phenomenon has been observed in numerous scientific areas such as hydrology, ethernet traffic data, stock returns and exchange rates, to name just a few. The extent of long memory dependence is characterized by the value of the so called Hurst exponent or Hurst coefficient H. Approximate normality and unbiasedness of the maximum likelihood estimate of H hold reasonably well for sample sizes as small as 20 if the mean and scale parameters are known. We show in a Monte Carlo study that if the latter two parameters are unknown, the bias and variance of the maximum likelihood estimate of H both increase substantially. We also show that the bias can be reduced by using a jackknife or parametric bootstrap proce- dure. However, in very large samples, maximum likelihood estimation becomes problematic because of the large dimension of the covariance matrix that must be inverted. We consider an approach for estimating the Hurst exponent by taking first order differ- ences of fractional Gaussian noise. We find that this differenced process has short memory and that, consequently, we may assume approximate independence between the estimates of the Hurst exponents in disjoint blocks of data. We split the data into a number of con- tiguous blocks, each containing a relatively small number of observations. Computation of the likelihood function in a block then presents no computational problem. We form a pseudo likelihood function consisting of the product of the likelihood functions in each of the blocks and provide a formula for the standard error of the resulting estimator of H. This formula is shown in a Monte Carlo study to provide a good approximation to the true standard error. Application of the methodology is illustrated in two data sets. The long memory property of a time series is primarily characterized by H. In general, such series are exceptionally long, therefore it is natural to enquire whether or not H remains constant over the full extent of the time series. We propose a number of tests for the hypothesis that H remains constant, against an alternative of a change in one or more values of H. Formulas are given to enable calculation of asymptotic p-values. We also propose a permutational procedure for evaluating exact p-values. The proposed tests are applied to three sets of data.