Dimensionality reduction in the bag-of-words vector space document representation model has been widely studied for the purposes of improving accuracy and reducing computational load of document retrieval tasks. These techniques, however, have not been studied to the same degree with regard to document clustering tasks. This study evaluates the effectiveness of two popular dimensionality reduction techniques for clustering, and their effect on discovering accurate and understandable topical groupings of documents. The two techniques studied are Latent Semantic Analysis and Independent Component Analysis, each of which have been shown to be effective in the past for retrieval purposes.

The discovery of gene regulatory network (GRN) using gene expression data is one of the promising directions for deciphering biological mechanisms, which underlie many basic aspects of scientific and medical advances. In this thesis, we focus on the reconstruction of GRN from time-series data using a Granger causality (GC) approach. As there is little existing research on combining data from multiple time-series experiments, we identify the need for developing a methodology with underlying theory to combine multiple experiments for statistical significant discovery. We derive a statistical theory for intersection of two discovered networks. Such a statistical framework is novel and intended for our GRN discovery problem. However, this theory is not limited to GRN or GC, and may be applied to other problems as long as one can take the intersection of discoveries obtained from multiple experiments (or datasets). We propose a number of novel methods for combining data from multiple experiments. Our single underlying model (SUM) method regresses data of multiple experiments in one go, enabling GC to fully utilize the information in the original data. Based on our statistical theory and SUM, we develop new meta-analysis methods, including union of pairwise common edges (UPCE) and leave-one-out hybrid of SUM and UPCE (LOOHSU). Applications on synthetic data and real data show that our new methods give discoveries of substantially higher precision than traditional meta-analysis. We also propose methods for estimating the precision of GC-discovered networks and thus fill in an important gap not considered in the literature. This allows us to assess how good a discovered network is in the case of unknown ground truth, which is typical in most biological applications. Our precision estimation by half-half splitting with combinations (HHSC) gives an estimate much closer to the true value compared with that computed from the Benjamini-Hochberg false discovery rate controlling procedure. Furthermore, using a network covering notion, we design a method that can identify a small number of links with high precision of around 0.8-0.9, which may relieve the burden of testing many hypothetical interactions of low precision in biological experiments. For the situation where the number of genes is much larger than the data length, in which case full-model GC cannot be applied, GC is often applied to the genes pairwisely. We analyze how spurious causalities (false discoveries) may arise. Consequently, we demonstrate that model validation can effectively remove spurious discoveries. With our proposed implementation that model orders are fixed by the Akaike information criterion and every model is subject to validation, we report a new observation that network hubs tend to act as sources rather than receivers of interactions. / published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

Background: Influenza viruses cause substantial mortality and morbidity both worldwide and in Hong Kong. Furthermore, the possible emergence of future influenza pandemics remains a major threat to public health. Some studies have estimated that one third of all influenza transmission occurs in households. Household studies have been an important means of studying influenza transmissions and evaluating the efficacy of influenza control measures including vaccination, antiviral therapy and prophylaxis and non-pharmaceutical interventions. Household studies of influenza can be categorized as pertaining to one of two designs: household cohort and case-ascertained. In household cohort studies households are recruited before the start of an influenza season and then monitored during the influenza season for influenza infection. In case-ascertained studies a household is enrolled once influenza infection is identified in a household member. Objectives: This thesis comprises of two parts. The objective of the first part is to evaluate the resource efficiency of different designs for conducting household studies. The objective of the second part is to estimate community and household transmission parameters during the 2009 A(H1N1) pandemic in Hong Kong. Methods: Monte Carlo simulation parameterized with data from influenza studies in Hong Kong was used to compare the resource efficiency of competing study designs evaluating the efficacy of an influenza control intervention. Approaches to ascertaining infections in different types of studies, and their implications for resource efficiency were compared. With regard to the second part, extended Longini-Koopman models within a Bayesian framework were used on data from a Hong Kong household cohort study conducted from December 2008 to October 2009. Household and community transmission parameters were estimated by age-groups for two seasonal influenza strains circulating in the winter of 2008-09 and two seasonal and one pandemic strain circulating in the summer of 2009. Results: Simulations showed that RT-PCR outperformed both serology and self-report of symptoms as a resource efficient means of identifying influenza in household studies. Identification of influenza using self-report of symptomatology performed particularly poorly in terms of resource efficiency due to its low sensitivity and specificity when compared to laboratory methods. Case-ascertained studies appeared more resource efficient than cohort studies but the results were sensitive to the choice of parameter values particularly the serial interval of influenza. In statistical analyses of household data during the winter of 2008-09, it was found that transmissibility of seasonal influenza strains were similar to those previously reported in the literature. Analysis also showed for the summer 2009 the estimates of household transmissibility were similar for seasonal A(H3N2) and pandemic A(H1N1) especially after taking into account that some individuals were likely immune to infection. Conclusions: Careful consideration of study design can ensure that studies are resource efficient and have sufficient statistical power. Data from a household study suggested that during 2009 seasonal and pandemic influenza had similar transmission patterns. / published_or_final_version / Community Medicine / Doctoral / Doctor of Philosophy

Not available / text

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

published_or_final_version / abstract / toc / Statistics and Actuarial Science / Master / Master of Philosophy

published_or_final_version / Statistics / Master / Master of Philosophy

published_or_final_version / Statistics and Actuarial Science / Master / Master of Philosophy

published_or_final_version / Statistics / Master / Master of Philosophy

No description available.