Finance should be studied as a hard science, where scientific methods apply. When a trading strategy is proposed, the underlying model should be transparent and defined robustly to allow other researchers to understand and examine it thoroughly. Any reports on experimental results must allow other researchers to trace back to the original data and models that produced them. Like any hard sciences, results must be repeatable to allow researchers to collaborate and build upon each other’s results. Large-scale collaboration, when applying the steps of scientific investigation, is an efficient way to leverage crowd science to accelerate research in finance. Unfortunately, the current reality is far from that. Evidence shows that current methods of investigation in finance in most cases do not allow for reproducible and falsifiable procedures of scientific investigation. As a consequence, the majority of financial decisions at all levels, from personal investment choices to overreaching global economic policies, rely on some variation of try-and-error and are mostly non-scientific by definition. We lack transparency for procedures and evidence, proper explanation of market events, predictability on effects, or identification of causes. There is no clear demarcation of what is inherently scientific, and as a consequence, the line between fake and true is blurred. In this research, we advocate the use of a next-generation investigative approach leveraging forces of human diversity, micro-specialized crowds, and proper computer-assisted control methods associated with accessibility, reproducibility, communication, and collaboration. This thesis is structured in three distinctive parts. The first part defines a set of very specific cognitive and non-cognitive enablers for crowd-based scientific investigation: methods of proof, large-scale collaboration, and a domain-specific computational representation. These enablers allow the application of procedures of structured scientific investigation powered by crowds, a “collective brain in which neurons are human collaborators”. The second part defines a specialized computational representation to allow proper controls and collaboration in large-scale in the field of economics. A computational representation is a role-based representation system based on facets, contributions, and constraints of data, and used to define concepts related to a specific domain of knowledge for crowd-based investigation. The third and last part performs an end-to-end investigation of a non-trivial problem in finance by measuring the actual performance of a momentum strategy in technical analysis, applying formal methods of investigation developed over the first and second part of this research.

Epidemiological data with disease characteristic information can be modelled in several ways. One way is taking each disease characteristic as a response and constructing binary or polytomous logistic regression model. Second way is using a new response which consists of disease subtypes created by cross-classification of disease characteristic levels, and then constructing polytomous logistic regression model. The former may be disadvantageous since any possible covariation between disease characteristics is neglected, whereas the latter can capture that covariation behaviour. However, cross-classifying the characteristic levels increases the number of categories of response, so that dimensionality problem in parameter space may occur in classical polytomous logistic regression model. A two staged polytomous logistic regression model overcomes that dimensionality problem. In this thesis, study is progressen in two main directions: simulation study and data analysis parts. In simulation study, models that capture the covariation behaviour are compared in terms of the response model parameter estimators. That is, performances of the maximum likelihood estimation (MLE) approach to classical polytomous logistic regression, Bayesian estimation approach to classical polytomous logistic regression and pseudo-conditional likelihood (PCL) estimation approach to two stage polytomous logistic regression are compared in terms of bias and variation of estimators. Results of the simulation study revealed that for small sized sample and small number of disease subtypes, PCL outperforms in terms of bias and variance. For medium scaled size of total disease subtypes situation when sample size is small, PCL performs better than MLE, however when the sample size gets larger MLE has better performance in terms of standard errors of estimates. In addition, sampling variance of PCL estimators of two stage model converges to asymptotic variance faster than the ML estimators of classical polytomous logistic regression model. In data analysis, etiologic heterogeneity in breast cancer subtypes of Turkish female cancer patients is investigated, and the superiority of the two stage polytomous logistic regression model over the classical polytomous logistic model with disease subtypes is represented in terms of the interpretation of parameters and convenience in hypothesis testing.

Bootstrapping is a resampling technique which treats the original data set as a population and draws samples from it with replacement. This technique is widely used, especially, in mathematically intractable problems. In this study, it is used to obtain the empirical distributions of the parameters to determine whether they are statistically significant or not in a special case of nonparametric regression, Conic Multivariate Adaptive Regression Splines (CMARS). Here, the CMARS method, which uses conic quadratic optimization, is a modified version of a well-known nonparametric regression model, Multivariate Adaptive Regression Splines (MARS). Although performing better with respect to several criteria, the CMARS model is more complex than that of MARS. To overcome this problem, and to improve the CMARS performance further, three different bootstrapping regression methods, namely, Random-X, Fixed-X and Wild Bootstrap are applied on four data sets with different size and scale. Then, the performances of the models are compared using various criteria including accuracy, precision, complexity, stability, robustness and efficiency. Random-X yields more precise, accurate and less complex models particularly for medium size and medium scale data even though it is the least efficient method.

This thesis presents an empirical investigation of the problems of including pictorial digital art in the context of Digital Libraries (DLs). The rational for this work is that digital art material is a significant source of learning and research, provided that it is systematically collected and maintained in structured electronic repositories. The thesis addresses a fundamental question: How to provide description and retrieval services, which are based on the needs of digital art user communities? This raises three research issues. One is the need to combine DL collections into meaningful and functional content. The second is the importance of a user-oriented approach to designing and developing Digital Libraries. The third is the requirement for continuing access to digital art as a record of modern culture. These questions are explored through a needs assessment targeted to Arts & Humanities scholars, digital artists and representatives of the DL community. A data collection methodology is developed, based on the principles of Social Informatics and a case study of evaluation efforts in extant projects. The results from this process demonstrate that the scholarly value of digital art can be established by aggregating material from various repositories into a unified dataset. The results also identify specific documentation and retrieval issues deriving from inclusion of digital art in a DL environment that necessitate further investigation. To this end, a review of sixteen digital art online resources is conducted which reveals ad-hoc collection strategies and metadata deficiencies. The work presents a prototype Digital Library for enhancing the educational outcome of digital art. The application is used as an implementation platform for material aggregation and augmented documentation through the Media Art Notation System (MANS). The summative evaluation findings confirm that the suggested solutions are highly rated by the targeted audiences. The thesis makes a contribution to academic knowledge in situating the representation of digital art within modern society. By critically examining the unique requirements of this material using the resources of social theory, the thesis represents a contemporary and pragmatic perspective on digital media art. In a well-structured Digital Library, the scholarly potential of digital art is much greater than the currently employed ad-hoc context. This work offers a sustained reflection and a roadmap for selecting and consistently applying a strategy that aims to continually improve the quality of digital art provision.

Significant global health attention and promotion has been focused on hand washing with soap due to the clear benefits observed in promoting and ensuring child health. However, the measurement and evaluation of hand washing behaviours remains complex. The Sanitation, Hygiene Education and Water Supply in Bangladesh Programme (SHEWA-B) is a large project being implemented by the Government of Bangladesh and UNICEF. This research assessed methodological issues of measuring hand washing behaviours through comparison of structured observation and responses to cross-sectional survey measures (spot-check observation, selfreported hand washing and a hand washing demonstration) and discusses the suitability of indicators. Focus group discussions with fieldworkers were also conducted. The results of this study indicate that hand washing behaviours were over-reported compared with structured observation findings. This implies that current estimates of hand washing from large scale surveys, for example, Demographic and Health Surveys (DHS) are also likely to be overestimates. In about 1000 households, approximately 1% or less of female caregivers were observed to wash their hands with soap or ash before preparing food, before eating, and 3% before feeding a child. Hand washing with soap was higher for defecation related events with approximately 29% of female caregivers using soap two thirds or more of the time after cleaning a child’s anus/disposing of a child’s stools and 38% used soap two-thirds or more of the time after defecation. Soap was observed at the hand washing location in about 50% of the households but actual practice was much lower. Reported knowledge was high; approximately 90% identified the important times for hand washing as being before eating and after defecation and approximately 50% identified before preparing food and after cleaning/changing a baby. The measurement of hand washing is complex and there has been limited research into the validity of different measurement methods. This research used an epidemiological style approach using the concepts of screening/diagnostic testing and calculation of kappa statistics to assess validity. In conclusion, this research demonstrates that self report hand washing measures are subject to over reporting. Structured observation provides useful information on directly observed hand washing behaviours and the frequency of behaviours. Spot check methods of soap and hand washing locations also provide more optimistic data than observations and can be used as an alternative to structured observation. In addition, the use of questions on the 24 hour recall of soap and other self report questions on knowledge and the availability of spare soap demonstrate potential for use as potential indicators as an alternative to structured observation. Further validation of measurement methods is required in different country settings.

This thesis explores the pathways through which individual and household factors are associated with temporal changes in child nutritional status. In this study the concept of nutrition deprivation is used in two ways: firstly as indicated by the child's anthropometric measures, and secondly in terms of food consumption. The thesis also explores how nutritional deprivation is linked with economic deprivation. The main objectives of the study are: to examine the physical growth trajectories of children, to investigate the household's economic and nutritional (food) deprivations, to explore the determinants of child malnutrition, and finally to investigate the relationship between temporal changes in the poverty status of households and temporal changes in child nutritional status. The study uses the Pakistan Panel Data collected by the International Food Policy Research Institute (IFPRI) from 1986-89, covering four rural districts of Pakistan. The study employs a comprehensive child health framework to establish the mechanism of child nutritional status by linking the various factors at child, household and community levels. This framework specifies poverty as the root cause of malnutrition. The basic need absolute poverty approach is used to work out the incidence and the dynamic nature of poverty. Various statistical modelling techniques for analysing the longitudinal data are used in this study. For example, to study the height and weight growth traectories of children a growth curve modelling technique is employed, and to study the determinants of child malnutrition a three-level hierarchical linear model for longitudinal data is used. The predicted average growth velocities indicate a slower growth during first year of child's life in comparison with the usual growth velocities amongst the normal children. However, in a particular cohort of children some evidence of growth acceleration is found during the third year of a child's life after a growth deceleration during the second year. Child level factors, such as breastfeeding and the incidence of diarrhoea and morbidity, are found to explain most of the variability in child nutritional status. The results reveal dissimilarities in nutritional status between children in a household. The results also indicate associations between poverty and stunting while chronic poverty is found to be associated with wasting. The results indicate that caloric and protein consumption amongst the study households was notably high. However, food consumption patterns mostly revolve around the staple food, and even in the top expenditure quintile this pattern remains persistent.

Background: The scope of postpartum pelvic dysfunction and perineal trauma is under-researched. Instrumental vaginal delivery or 3rd/4th degree tears were recognised risk factors for pelvic/perineal dysfunction; caesarean delivery was not implicated. Aims: • To analyse obstetrical/biological factors associated with pelvic dysfunction after caesarean or non-instrumental vaginal delivery • To compare these associations between groups after determining frequencies • To evaluate severity of pelvic/perineal dysfunction, including quantifying maternal perception of the psychosocial impact Participants and Methods: 284 primiparae (184 caesarean, 100 vaginally delivered) had domiciliary, in-depth medical interviews using structured and open questioning. Results: Caesarean (elective, emergency) vs. vaginally delivered were compared: Stress incontinence manifested in 60/184 (33%, 33%) vs. 54/100 (54%), anal incontinence in 94/184 (53%, 50%) vs. 44/100 (44%), dyspareunia in 50/184 (28%, 27%) vs. 46/100 (46%), haemorrhoids in 3/184 (2%) vs. 5/100 (5%) and double incontinence with dyspareunia in 33/284 (14%, 10% vs. 12%). Sixty sustained perineal trauma. Delivery mode and non-labour factors were predictors. Severity was evaluated by devising a psychosocial measure tailored to maternal functioning. New faecal incontinence necessitated continuous perineal protection in two pre-labour caesarean and one vaginally delivered mother. Severe dysphoria was associated with incontinence (p=0.038, OR 2.334, CI 1.049, 5.192), dyspareunia (p=0.005, OR 2.231, CI 1.272, 3.914) and post-caesarean wound problems (p=0.022, OR 3.620, CI 1.203, 10.896). Incontinence impaired leisure activities (p=0.036, OR 2.165, CI 1.051, 4.463) and employment (p=0.023, OR 1.912, CI 1.093, 3.345); caesarean mode affected social-networking (p=0.018, OR 2.438, CI 1.166, 5.099) and employment (p=0.031, OR 1.967, CI 1.064, 3.636). Conclusions: Pelvic/perineal dysfunction was: ▪ Predicted by caesarean or non-instrumental vaginal delivery, with anal incontinence being more prevalent post-caesarean ▪ Comparable following elective or emergency caesarean ▪ Associated with severe and quantifiable biopsychosocial maternal morbidity.

ABSTRACT ROBUST CONTROL CHARTS &Ccedil / etiny&uuml / rek, Aysun M. Sc., Department of Statistics Supervisor: Dr. BariS S&uuml / r&uuml / c&uuml / Co-Supervisor: Assoc. Prof. Dr. Birdal Senoglu December 2006, 82 pages Control charts are one of the most commonly used tools in statistical process control. A prominent feature of the statistical process control is the Shewhart control chart that depends on the assumption of normality. However, violations of underlying normality assumption are common in practice. For this reason, control charts for symmetric distributions for both long- and short-tailed distributions are constructed by using least squares estimators and the robust estimators -modified maximum likelihood, trim, MAD and wave. In order to evaluate the performance of the charts under the assumed distribution and investigate robustness properties, the probability of plotting outside the control limits is calculated via Monte Carlo simulation technique.

The longitudinal feature of measurements and counting process of responses motivate the regression models for longitudinal count data (LCD) to take into account the phenomenons such as within-subject association and overdispersion. One common problem in longitudinal studies is the missing data problem, which adds additional difficulties into the analysis. The missingness can be handled with missing data techniques. However, the amount of missingness in the data and the missingness mechanism that the data have affect the performance of missing data techniques. In this thesis, among the regression models for LCD, the Log-Log-Gamma marginalized multilevel model (Log-Log-Gamma MMM) and the random-intercept model are focused on. The performance of the models is compared via a simulation study under three missing data mechanisms (missing completely at random, missing at random conditional on observed data, and missing not random), two types of missingness percentage (10% and 20%), and four missing data techniques (complete case analysis, subject, occasion and conditional mean imputation). The simulation study shows that while the mean absolute error and mean square error values of Log-Log-Gamma MMM are larger in amount compared to the random-intercept model, both regression models yield parallel results. The simulation study results justify that the amount of missingness in the data and that the missingness mechanism that the data have, strictly influence the performance of missing data techniques under both regression models. Furthermore, while generally occasion mean imputation displays the worst performance, conditional mean imputation shows a superior performance over occasion and subject mean imputation and gives parallel results with complete case analysis.

Conditional Cash Transfer (CCT) is a social assistance program which aims for investing in human capital by enabling families under risk of poverty to send their children to school and to benefit from health services regularly. CCT aims for decreasing poverty by means of cash transfers in the short run and aims for investing in children&rsquo / s human capital by providing basic preventative health care, regular school attendance and nutrition in the long run. Under the state of these aims, beginning from 1990s, more than 20 countries in the world started their own CCT program by the mediation or leadership of World Bank. CCT program in Turkey started so as to decrease the adverse effects of economic crisis in 2001 within the Social Risk Mitigation Project which was financially supported by the World Bank loan and constituted under the Social Assistance and Solidarity Foundation. CCT program in Turkey has been adopted by poor families in recent years, and demands and overall payments within the program have been increased significantly in a consideration of years. The need for examining and predicting the increase in these demands scientifically / and considering the fact that CCT is being applied over 20 countries, and such a study being never done before made this study necessary. In this thesis study, the change of CCT applications and number of beneficiary household over time were modeled using multivariate time series models according to geographical regions. Using the vector autoregressive models with exogenous variables (VARX), the forecasts were obtained for the number of CCT applications and beneficiary households in the future.