Approved for public release, distribution is unlimited / The J-52 engine used in the EA-6B Prowler has been found to have a faulty design which has led to in-flight engine failures due to the degradation of the 4.5 roller bearing. Because of cost constraints, the Navy developed a policy of maintaining rather than replacing the faulty engine with a re-designed engine. With an increase in Prowler crashes related to the failure of this bearing, the Navy has begun to re-evaluate this policy. This thesis analyzed the problem using methods in reliability statistics to develop policy recommendations for the Navy. One method analyzed the individual times to failure of the bearings and fit the data to a known distribution. Using this distribution, we estimated lower confidence bounds for the time which 0.0001% of the bearings are expected to fail, finding it was below fifty hours. Such calculations can be used to form maintenance and replacement policies. Another approach analyzed oil samples taken from the J-52 engine. The oil samples contain particles of different metals that compose the 4.5 roller bearing. Linear regression, classification and regression trees, and discriminant analysis were used to determine that molybdenum and vanadium levels are good indicators of when a bearing is near failure. / http://hdl.handle.net/10945/945 / Ensign, United States Navy

Doctor of Philosophy / Department of Civil Engineering / Sunanda Dissanayake / Malgorzata J. Rys / Driver workload increases on horizontal curves due to more complicated navigation compared to navigation on straight roadway sections. Although only a small portion of roadways are horizontal curve sections, approximately 25% of all fatal highway crashes occur at horizontal curve sections. According to the Fatality Analysis Reporting System (FARS) database, fatalities associated with horizontal curves were more than 25% during last years from 2008 to 2014, reinforcing that investigation of horizontal curve crashes and corresponding safety improvements are crucial study topics within the field of transportation safety. Improved safety of horizontal curve sections of rural transportation networks can contribute to reduced crash severities and frequencies. Statistical methods can be utilized to develop crash prediction models in order to estimate crashes at horizontal curves and identify contributing factors to crash occurrences, thereby correlating to the primary objectives of this research project. Primary data analysis for 221 randomly selected horizontal curves on undivided two-lane two-way highways with Poisson regression method revealed that annual average daily traffic (AADT), heavy vehicle percentage, degree of curvature, and difference between posted and advisory speeds affect crash occurrence at horizontal curves. The data, however, were relatively overdispersed, so the negative binomial (NB) regression method was utilized. Results indicated that AADT, heavy vehicle percentage, degree of curvature, and long tangent length significantly affect crash occurrence at horizontal curve sections. A new dataset consisted of geometric and traffic data of 5,334 horizontal curves on the entire state transportation network including undivided and divided highways provided by Kansas Department of Transportation (KDOT) Traffic Safety Section as well as crash data from the Kansas Crash and Analysis Reporting System (KCARS) database were used to analyze the single vehicle (SV) crashes. An R software package was used to write a code and combine required information from aforementioned databases and create the dataset for 5,334 horizontal curves on the entire state transportation network. Eighty percent of crashes including 4,267 horizontal curves were randomly selected for data analysis and remaining 20% horizontal curves (1,067 curves) were used for data validation. Since the results of the Poisson regression model showed overdispersion of crash data and many horizontal curves had zero crashes during the study period from 2010 to 2014, NB, zero-inflated Poisson (ZIP), and zero-inflated negative binomial (ZINB) methods were used for data analysis. Total number of crashes and severe crashes were analyzed with the selected methods. Results of data analysis revealed that AADT, heavy vehicle percentage, curve length, degree of curvature, posted speed, difference between posted and advisory speed, and international roughness index influenced single vehicle crashes at 4,267 randomly selected horizontal curves for data analysis. Also, AADT, degree of curvature, heavy vehicle percentage, posted speed, being a divided roadway, difference between posted and advisory speeds, and shoulder width significantly influenced severe crash occurrence at selected horizontal curves. The goodness-of-fit criteria showed that the ZINB model more accurately predicted crash numbers for all crash groups at the selected horizontal curve sections. A total of 1,067 horizontal curves were used for data validation, and the observed and predicted crashes were compared for all crash groups and data analysis methods. Results of data validation showed that ZINB models for total crashes and severe crashes more accurately predicted crashes at horizontal curves. This study also investigated the effect of speed limit change on horizontal curve crashes on K-5 highway in Leavenworth County, Kansas. A statistical t-test proved that crash data from years 2006 to 2012 showed only significant reduction in equivalent property damage only (EPDO) crash rate for adverse weather condition at 5% significance level due to speed limit reduction in June 2009. However, the changes in vehicles speeds after speed limit change and other information such as changes in surface pavement condition were not available. According to the results of data analysis for 221 selected horizontal curves on undivided two-lane highways, tangent section length significantly influenced total number of crashes. Therefore, providing more information about upcoming changes in horizontal alignment of the roadway via doubling up warning sings, using bigger sings, using materials with higher retroreflectivity, or flashing beacons were recommended for horizontal curves with long tangent section lengths and high number of crashes. Also, presence of rumble strips and wider shoulders significantly and negatively influenced severe SV crashes at horizontal curve sections; therefore, implementing rumble strips and widening shoulders for horizontal curves with high number of severe SV crashes were recommended.

MSc., Faculty of Science, University of the Witwatersrand, 2011 / The study investigates models and processes that could be used by decision-makers to influence policy. The objectives of the study are to illustrate the considerations in the use of information systems and modeling in the context of sustainable development in South Africa. The problem considered is that of measuring South Africa‟s progress toward sustainable development. The study is conducted at a national level. The design uses statistical indicators arranged into a model based information system using the Bellagio principles of 1996. The statistical indicators are assembled from the United Nations 1993 System of National Accounts, UN 1995 Driver-Pressure-State-Impact-Response framework, United Nations 2003 System of integrated Environmental and Economic Accounting, Basel Committee‟s 2004 Basel II framework, and South African Presidency 2010 Developmental indicators frameworks. The South African economy is conceptualised as a dynamic system composed of five types of capital. The fitted model is a vector autoregressive time series model of order p on a set of statistical factors that describe the South African economy. The robustness of the model to assumptions is evaluated using sensitivity and uncertainty analysis. The optimality of the model output for decision making is evaluated using decision theory. The study will facilitate an evidence based approach to managing South Africa‟s progress towards sustainable development.

Since bias threatens the validity of a study, it should be avoided where possible. Across all phases of a research project, bias could be introduced, and in most situations the researcher has reasonable control over processes that may be the source of bias. However, within a quantitative research context in social sciences, where the opinions, attitudes and intentions of people are often sought, response styles patterns due to cultural background, for example, are not within the control of the researcher. Typical response style patterns include acquiescence bias, a tendency to be agreeable to statements, which could be more prevalent in certain cultural groups than other. Another response style pattern is extremity ratings, where respondents tend to avoid the middle categories and mark the scale extremes. When practitioners sample respondents from different cultural groups, it is difficult, and depending on the research design, sometimes impossible to know whether significant differences are an artefact of substantive differences, or of differences in response styles. Adjusting scores for bias has a significant effect on the interpretation of research findings. To correct for bias, the method most commonly used to adjust scores within each cultural group is standardisation. In this research, SIMNORM, a target distribution estimation approach was used for the simultaneous estimation of a class of non-linear transformation functions that transform the composite scores within each cultural group to a standard normal distribution. SIMNORM was found to perform better than standardisation to obtain equivalence across cultural groups when composite scores are used. In addition, SIMITNORM, an item normalisation approach was developed, which is a simultaneous non-linear transformation of item scores to a standard normal target distribution. The results of seven nested SIMITNORM models were compared to raw item scores and standardised scores, using a multi-group confirmatory factor analysis approach, a method that is suitable to test for construct equivalence, metric equivalence and scalar equivalence. SIMITNORM had significant advantages over standardisation as an approach to obtain equivalence over items in a set of data where bias is present.

A Research Report submitted in partial fulfilment of the requirements for the degree of Master of Science in Engineering (Mining) to the Faculty of Engineering and the Built Environment, University of the Witwatersrand, Johannesburg, South Africa, July 2017 / The project investigated the application of geostatistical techniques in evaluating a mechanically deposited platinum tailings resource. The project was undertaken on one of the Anglo American Platinum tailings dams, the identity of which cannot be revealed, due to the agreement in place or permission given. Remnant unrecovered minerals of economic potential still exist in tailings dams. These unrecovered minerals have influenced several mining companies to turn their attention to the economic potential that still exists in tailings, making them a key strategic component of their resources and reserves. Geostatistics has been developed and thoroughly tested or improved to address challenges experienced in estimating in situ geological ore bodies. The main aim of this Research Project is to test whether these fundamental principles and theories of geostatistics are relevant and appropriate in evaluating man-made ore bodies, such as a Platinum tailings dam, without any significant changes needed on the underlying principles or estimation algorithms. The findings on the Case Study tailings resource can be applied in the evaluation of other tailings dams, as well as any other man-made structures such as low grade rock dumps, muck piles, with related characteristics. A standard approach (methodology) was followed to evaluate the Case Study tailings resource. Drilling and sampling was conducted through sonic drilling. It is a dry drilling technique that is suitable for sampling unconsolidated particles such as tailings. Thereafter, 2 samples were sent to the laboratory to establish grade (concentration) of Platinum Group Metals (Platinum, Palladium and Rhodium), Gold and Base Metals (Copper and Nickel). Density was also measured, and comprehensively analysed as part of variables of interest in this research. Statistical analyses were performed on all variables of interest contained in the dam: which are Platinum (Pt), Palladium (Pd), Gold (Au), 3E (two PGMs plus Gold), Copper (Cu), Nickel (Ni) and Density. The underlying statistical distributions of all metals and density were found to be non-symmetrical and slightly positive skewed. The skewness of the distributions was established to be marginal. Differences between raw data (untransformed) averages and the log-normal estimates were analysed and found to be insignificant. As such Ordinary Kriging of untransformed data was concluded to be the appropriate geostatistical technique for Case Study tailings resource. Analysis of mineralisation continuity (variography), a pre-requisite for geostatistical techniques such as Ordinary Kriging applied on the case study tailings resource, was also performed. Reasonable and sufficient mineralisation continuity was established to exist in the Case Study tailings resource. Although characterised by high nugget effect, these spatial correlations were established to be continuous with ranges of influence well beyond 450 m in all variables. Anisotropic variograms were modelled for all variables and are comprised of nested structures with two to three spherical models. Resource estimation was conducted through Ordinary Kriging in Datamine. All the seven variables were successfully interpolated into each cell of the 5m x 5m x 5m block model. Rigorous validation of the resource model was performed to establish the quality and reliability of the estimation carried out. Estimated resource model was analysed against the original borehole data, through comparison of grade profiles, statistical analysis, QQ Plots and histograms. The grade profile was recognised to be similar between boreholes (5 m composites) and the adjacent cells that have been estimated. Furthermore, statistical analyses revealed minimal differences between means of the estimated model and the original borehole data: the highest difference being 1.7% realised on 3E, followed by 1.1% on Density and Gold (Au). The rest of the variables (Pt, Pd, Cu, and Ni) have differences that are below 1%. 3 QQ plots and histogram were plotted from resource model with 5m x 5m x 5m cells and 5 m composited boreholes. Although these data sets are of different (slightly incompatible) supports, the intended purpose of comparing distributions was achieved. QQ plots and histograms revealed approximately identical shaped distributions of the two data sets, with some minor deviations noticeable in graphs of only two variables (Au and Density) that are underlain by two populations. The validation process carried out gave a compelling assurance on the quality and reliability of the resource model produced. The Case Study tailings resource therefore is successfully estimated by Ordinary Kriging. The results achieved on the Case Study tailings dam has successfully proved that geostatistical principles and theories can confidently be applied, in their current form or understanding, to any man-made tailings resource / CK2018

Thesis (M.Sc.(Engineering))--University of the Witwatersrand, Faculty of Engineering, 1951.

As curvas de Rosso e de Siqueira para avaliação do estado nutricional de gestantes foram comparadas, em uma amostra de 565 gestantes. Para que essa comparação fosse feita a contento, utilizaram-se técnicas estatísticas tradicionais como: regressão múltipla por stepwise, coeficientes de associação, de concordância, de especificidade e de sensibilidade, métodos gráficos e técnicas mais atuais de classificação, como logística multinomial, regressão e classificação por árvores do tipo AID. Os resultados indicaram vantagem da curva de Siqueira, em termos de consistência, eficiência e associação, com relação ao prognóstico para a adequação do peso do RN. A análise das interrelações mostraram que a curva de Rosso tem uma tendência em sub-classificar a gestante, com relação ao seu peso e ao peso do seu RN. Para validar e confirmar tais resultados, procurou-se comparar as técnicas de classificação, uma vez que existiam cinco técnicas, cada qual, devido às suas características, com alguns resultados distintos. A comparação entre as técnicas de classificação foi feita por intermédio da taxa de má- classificação, usando validação cruzada em 1 O e 5 partes, obtendo-se vantagem das técnicas CHAID e Logística Multinomial, seguidas de CART, SPLUS, KS. A validação cruzada permitiu também confirmar a vantagem da curva de Siqueira, com o destaque para o fato que o melhor diagnóstico, com relação ao prognóstico da adequação do peso do RN, é na primeira consulta do pré-natal, abrindo o horizonte de diagnóstico precoce, prevenção e intervenção logo no início da gravidez, o que indica o seu uso com maiores chances de bons resultados. / The curves of Rosso and Siqueira, used to evaluate the nutritional state of pregnant women, were compared in a sample of 565 mothers. In order to make this comparison sucessful, traditional statistical techniques were used such as: multiple regression by stepwise, coefficients of association, concordance, specificity and sensibility, and graphical methods, and the updated techniques of classification, such as: multinominal logistic, regression and classification through trees of AID type. The results showed the advantage of the Siqueira curve, in terms of consistency, eficiency and association, related to prognosis of the adequacy of the weight of the newbom. The analysis of the interrelations showed that the curve of Rosso has a tendency to under classify the pregnant women concerning their weight and the weight of the newborn. In order to validate and confirm these results, comparisons of classification of techniques were searched, since there were five techniques and, each one, due to its own characteristics, presented some distinct results. The comparison between the techniques of classification were made by means of misclassification rate, using cross-validation in 10 and 5 parts, where CHAID techniques and Multinominal Logistic showed advantages, followed by CART, SPLUS and KS. The cross-validation also allowed us confirm the advantage of the Siqueira curve, with emphasis on the fact that the best diagnose, concerning the prognosis of the adequacy of the weight of the newborn, is obtained in the first antenatal care appointment. This expands horizons for precocious diagnose, prevention and intervention at the very beginning of pregnancy, which indicates its use with better chances of good results.

As curvas de Rosso e de Siqueira para avaliação do estado nutricional de gestantes foram comparadas, em uma amostra de 565 gestantes. Para que essa comparação fosse feita a contento, utilizaram-se técnicas estatísticas tradicionais como: regressão múltipla por stepwise, coeficientes de associação, de concordância, de especificidade e de sensibilidade, métodos gráficos e técnicas mais atuais de classificação, como logística multinomial, regressão e classificação por árvores do tipo AID. Os resultados indicaram vantagem da curva de Siqueira, em termos de consistência, eficiência e associação, com relação ao prognóstico para a adequação do peso do RN. A análise das interrelações mostraram que a curva de Rosso tem uma tendência em sub-classificar a gestante, com relação ao seu peso e ao peso do seu RN. Para validar e confirmar tais resultados, procurou-se comparar as técnicas de classificação, uma vez que existiam cinco técnicas, cada qual, devido às suas características, com alguns resultados distintos. A comparação entre as técnicas de classificação foi feita por intermédio da taxa de má- classificação, usando validação cruzada em 1 O e 5 partes, obtendo-se vantagem das técnicas CHAID e Logística Multinomial, seguidas de CART, SPLUS, KS. A validação cruzada permitiu também confirmar a vantagem da curva de Siqueira, com o destaque para o fato que o melhor diagnóstico, com relação ao prognóstico da adequação do peso do RN, é na primeira consulta do pré-natal, abrindo o horizonte de diagnóstico precoce, prevenção e intervenção logo no início da gravidez, o que indica o seu uso com maiores chances de bons resultados. / The curves of Rosso and Siqueira, used to evaluate the nutritional state of pregnant women, were compared in a sample of 565 mothers. In order to make this comparison sucessful, traditional statistical techniques were used such as: multiple regression by stepwise, coefficients of association, concordance, specificity and sensibility, and graphical methods, and the updated techniques of classification, such as: multinominal logistic, regression and classification through trees of AID type. The results showed the advantage of the Siqueira curve, in terms of consistency, eficiency and association, related to prognosis of the adequacy of the weight of the newbom. The analysis of the interrelations showed that the curve of Rosso has a tendency to under classify the pregnant women concerning their weight and the weight of the newborn. In order to validate and confirm these results, comparisons of classification of techniques were searched, since there were five techniques and, each one, due to its own characteristics, presented some distinct results. The comparison between the techniques of classification were made by means of misclassification rate, using cross-validation in 10 and 5 parts, where CHAID techniques and Multinominal Logistic showed advantages, followed by CART, SPLUS and KS. The cross-validation also allowed us confirm the advantage of the Siqueira curve, with emphasis on the fact that the best diagnose, concerning the prognosis of the adequacy of the weight of the newborn, is obtained in the first antenatal care appointment. This expands horizons for precocious diagnose, prevention and intervention at the very beginning of pregnancy, which indicates its use with better chances of good results.

Lim Sze Wah. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2002. / Includes bibliographical references (leaves 71-73). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 2 --- A brief review on the basic minority game --- p.6 / Chapter 2.1 --- Model --- p.7 / Chapter 2.2 --- Results --- p.9 / Chapter 2.3 --- Discussion --- p.11 / Chapter 2.3.1 --- Bit-string Statistics and Market Efficiency --- p.11 / Chapter 2.3.2 --- Crowds and Anticrowds Effect --- p.14 / Chapter 2.3.3 --- Hamming Distance and Reduced Strategies Space --- p.15 / Chapter 3 --- A brief review on existing variations on the minority game --- p.17 / Chapter 3.1 --- Darwinism process and MG --- p.17 / Chapter 3.2 --- Evolutionary MG (EMG) --- p.17 / Chapter 3.3 --- Modified EMG (MEMG) --- p.18 / Chapter 3.4 --- MG with arbitrary cutoff --- p.18 / Chapter 3.5 --- Thermal MG (TMG) --- p.19 / Chapter 3.6 --- Three-Sided MG --- p.19 / Chapter 3.7 --- MG with variable payoffs --- p.19 / Chapter 4 --- Minority game with varying number of participants --- p.21 / Chapter 4.1 --- The modified MG --- p.22 / Chapter 4.1.1 --- Model --- p.22 / Chapter 4.1.2 --- Results --- p.23 / Chapter 4.2 --- Mixed-population --- p.33 / Chapter 4.2.1 --- Model --- p.33 / Chapter 4.2.2 --- Results --- p.33 / Chapter 4.2.3 --- Discussions --- p.37 / Chapter 5 --- Minority game considering recent performance of strategies --- p.39 / Chapter 5.1 --- The modified MG --- p.40 / Chapter 5.1.1 --- Model --- p.40 / Chapter 5.1.2 --- Results --- p.40 / Chapter 5.2 --- Mixed-population --- p.46 / Chapter 5.2.1 --- Model --- p.46 / Chapter 5.2.2 --- Results --- p.47 / Chapter 5.2.3 --- Discussions --- p.50 / Chapter 6 --- Minority game combining both modifications --- p.52 / Chapter 6.1 --- Model --- p.52 / Chapter 6.2 --- Results --- p.53 / Chapter 7 --- Conclusion --- p.68 / Bibliography --- p.71

Lam Yuk Hing. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2002. / Includes bibliographical references (leaves 67-70). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 2 --- Diagnostic Measure --- p.7 / Chapter 2.1 --- A diagnostic measure di based on Cook's likelihood distance --- p.7 / Chapter 2.2 --- The estimates 6 and θ (i) --- p.8 / Chapter 2.3 --- "The one-step estimator,θ1 (i)" --- p.9 / Chapter 3 --- Methods For Identifying Influential Observations --- p.11 / Chapter 3.1 --- One-step method --- p.11 / Chapter 3.2 --- Forward search procedure --- p.13 / Chapter 3.2.1 --- Idea of forward search procedure --- p.14 / Chapter 3.2.2 --- "The modified diagnostic measure, di" --- p.15 / Chapter 3.2.3 --- Initial basic subset --- p.18 / Chapter 3.2.4 --- The algorithm of starting with an ordered basic subset --- p.19 / Chapter 3.2.5 --- The algorithm of starting with a random basic subset --- p.21 / Chapter 4 --- Case Study --- p.23 / Chapter 4.1 --- Open/Close Book data set --- p.23 / Chapter 4.1.1 --- One-step method --- p.27 / Chapter 4.1.2 --- Forward search procedure --- p.28 / Chapter 4.1.3 --- Start with the ordered basic subset --- p.28 / Chapter 4.1.4 --- Start with a random basic subset --- p.32 / Chapter 4.2 --- Paper-Quality Measurements data set --- p.38 / Chapter 4.2.1 --- One-step method --- p.40 / Chapter 4.2.2 --- Forward search procedure --- p.41 / Chapter 4.2.3 --- Start with the ordered basic subset --- p.41 / Chapter 4.2.4 --- Start with a random basic subset --- p.45 / Chapter 5 --- Simulation --- p.52 / Chapter 5.1 --- Simulation procedure --- p.52 / Chapter 5.2 --- Results --- p.56 / Chapter 6 --- Discussion --- p.64 / Reference --- p.67