Indiana University-Purdue University Indianapolis (IUPUI) / Context. Prediction models for dementia are crucial for informing clinical decision making in older adults. Previous models have used genotype and age to obtain risk scores to determine risk of Alzheimer’s Disease, one of the most common forms of dementia (Desikan et al., 2017). However, previous prediction models do not account for the fact that the time to dementia onset is unknown, lying between the last negative and the first positive dementia diagnosis time (interval censoring). Instead, these models use time to diagnosis, which is greater than or equal to the true dementia onset time. Furthermore, these models do not account for the competing risk of death which is quite frequent among elder adults. Objectives. To develop a prediction model for dementia that accounts for interval censoring and the competing risk of death. To compare the predictions from this model with the predictions from a naïve analysis that ignores interval censoring and the competing risk of death. Methods. We apply the semiparametric sieve maximum likelihood (SML) approach to simultaneously model the cumulative incidence function (CIF) of dementia and death while accounting for interval censoring (Bakoyannis, Yu, & Yiannoutsos, 2017). The SML is implemented using the R package intccr. The CIF curves of dementia are compared for the SML and the naïve approach using a dataset from the Indianapolis Ibadan Dementia Project. Results. The CIF from the SML and the naïve approach illustrated that for healthier individuals at baseline, the naïve approach underestimated the incidence of dementia compared to the SML, as a result of interval censoring. Individuals with a poorer health condition at baseline have a CIF that appears to be overestimated in the naïve approach. This is due to older individuals with poor health conditions having an elevated risk of death. Conclusions. The SML method that accounts for the competing risk of death along with interval censoring should be used for fitting prediction/prognostic models of dementia to inform clinical decision making in older adults. Without controlling for the competing risk of death and interval censoring, the current models can provide invalid predictions of the CIF of dementia.

To improve the safety and comfort of intelligent vehicles, advanced driver models offer promising solutions. However, several shortcomings of these models prevent them from being widely applied in reality. To address these shortcomings, advanced artificial intelligence algorithms in conjunction with the sufficient driving environmental factors are proposed based on real-life driving data. More specifically, three typical problems will be addressed in this thesis: Mandatory Lane Changing (MLC) suggestion at the highway entrance; Discretionary Lane Changing (DLC) intention prediction; Car-Following gap model considering the effect of cuts-in from the adjacent lanes. For the MLC suggestion system, in which the main challenges are efficient decision making and high prediction accuracy of both non-merge and merge events, an additional gated branch neural network (GBNN) is proposed. The proposed GBNN algorithm not only achieves the highest accuracy among conventional binary classifiers in terms of great performance on the non-merge accuracy, the merge accuracy, and receiver operating characteristic score but also takes less time. For the DLC, we propose a recurrent neural network (RNN)-based time series classifier with a gated recurrent units (GRU) architecture to predict the surrounding vehicles’ intention. It can predict the surrounding vehicles’ lane changing maneuver 0.8 s in advance at a recall and precision of 99.5% and 98.7%, respectively, which outperforms conventional algorithms such as the Hidden Markov Model (HMM). Finally, drivers are typically faced with two competing challenges when following a preceding vehicle. A method is proposed to address the problem through an overall objective function of car-following gap and velocity. Based on this, seeking the strategic car-following gap translates to finding the optimal solution that minimizes the overall objective function. With the support of field data, the method along with concrete models are instantiated and the application of the method is elaborated. / Thesis / Doctor of Philosophy (PhD) / Lane changing and car following are the two most frequently encountered driving behaviours for intelligent vehicles. Substantial research has been carried out and several prototypes have been developed by universities as well as companies. However, the low accuracy and high computational cost prevent the existing lane changing models from providing safer and more reliable decisions for intelligent vehicles. In the existing car-following models, there are also few models that consider the effects of cut-ins from adjacent lanes which may result in their poor accuracy and efficiency. To address these obstacles, advanced artificial intelligence algorithms combined with sufficient driving environmental factors are proposed due to their promise of providing accurate, efficient, and robust lane changing and car-following models. The main part of this thesis is composed of three journal papers. Paper 1 proposed a gated branch neural network for a mandatory lane changing suggestion system at the on-ramps of highways; paper 2 developed a recurrent neural network time-series algorithm to predict the surrounding vehicles’ discretionary lane changing intention in advance; paper 3 researched the strategic car-following gap model considering the effect of cut-ins from adjacent lanes.

Among natural disasters affecting the United States, North Atlantic tropical cyclones (TCs) and hurricanes are responsible for the highest economic losses and are one of the main causes of fatalities. Although we cannot prevent these storms from occurring, skillful seasonal predictions of the North Atlantic TC activity and associated impacts can provide basic information critical to our improved preparedness. Unfortunately, it is not yet possible to predict heavy rainfall and flooding associated with these storms several months in advance, and the lead time is limited to few days at the most. On the other hand, overall North Atlantic TC activity can be potentially predicted with a six- to nine-month lead time. This thesis focuses on the evaluation of the skill in predicting basin-wide North Atlantic TC activity with a long lead time and rainfall with a short lead time. For the seasonal forecast of TC activity, we develop statistical-dynamical forecasting systems for different quantities related to the frequency and intensity of North Atlantic TCs using only tropical Atlantic and tropical mean sea surface temperatures (SSTs) as covariates. Our results show that skillful predictions of North Atlantic TC activity are possible starting from November for a TC season that peaks in the August-October months. The short term forecasting of rainfall associated with TC activity is based on five numerical weather prediction (NWP) models. Our analyses focused on 15 North Atlantic TCs that made landfall along the U.S. coast over the period of 2007-2012. The skill of the NWP models is quantified by visual examination of the distribution of the errors for the different lead-times, and numerical examination of the first three moments of the error distribution. Based on our results, we conclude that the NWP models can provide skillful forecasts of TC rainfall with lead times up to 48 hours, without a consistently best or worst NWP model.

Warfarin is one of the most widely used anticoagulants in the world. Treatment is complicated by a large inter-individual variation in the dose needed to reach adequate levels of anticoagulation i.e. INR 2.0 – 3.0. The objective of this thesis was to evaluate which factors, mainly genetic but also non-genetic, that affect the response to warfarin in terms of required maintenance dose, efficacy and safety with special focus on warfarin dose prediction. Through candidate gene and genome-wide studies, we have shown that the genes CYP2C9 and VKORC1 are the major determinants of warfarin maintenance dose. By combining the SNPs CYP2C9 *2, CYP2C9 *3 and VKORC1 rs9923231 with the clinical factors age, height, weight, ethnicity, amiodarone and use of inducers (carbamazepine, phenytoin or rifampicin) into a prediction model (the IWPC model) we can explain 43 % to 51 % of the variation in warfarin maintenance dose. Patients requiring doses < 29 mg/week and doses ≥ 49 mg/week benefitted the most from pharmacogenetic dosing. Further, we have shown that the difference across ethnicities in percent variance explained by VKORC1 was largely accounted for by the allele frequency of rs9923231. Other novel genes affecting maintenance dose (NEDD4 and DDHD1), as well as the replicated CYP4F2 gene, have small effects on dose predictions and are not likely to be cost-effective, unless inexpensive genotyping is available. Three types of prediction models for warfarin dosing exist: maintenance dose models, loading dose models and dose revision models. The combination of these three models is currently being used in the warfarin treatment arm of the European Pharmacogenetics of Anticoagulant Therapy (EU-PACT) study. Other clinical trials aiming to prove the clinical validity and utility of pharmacogenetic dosing are also underway. The future of pharmacogenetic warfarin dosing relies on results from these ongoing studies, the availability of inexpensive genotyping and the cost-effectiveness of pharmacogenetic driven warfarin dosing compared with new oral anticoagulant drugs.

Melanin pigmentation is a complex trait governed by many genes. Variation in melanin pigmentation within, and between, populations makes it an important trait for assisting in physical identification of an individual in forensic investigations. Utilizing a training sample (n=789) comprised of various ethnicities and SNPs (75) in 24 genes previously implicated in human or animal pigmentation studies, I determined three-SNP multiple linear regression models that accounted for large proportions of pigmentation variation in skin (45.7%), eye color (76.4%), and hair [eumelanin-to-pheomelanin (43.2%) and total melanin (76.3%)], independent of ethnic origin. Rather than implementing stepwise regression, to ascertain the three-SNP predictive models, I devised an algorithm that is likely more robust than stepwise regression. The algorithm consisted of two steps: the first step reduced the pool of 75 SNPs to a pool of 40 by selection of SNPs that were significant (p<0.05) by one-way ANOVA; the second step enabled selection of SNPs for model incorporation based on their frequency in the best-fitted models of all possible combinations of three-SNP models (i.e., 40 choose 3).Prediction models were validated utilizing an independent cohort (n=242, test sample) that was very similar in ethnic composition to the training sample. Relative shrinkage was moderate for skin reflectance (23.4%), eye color (19.4%), and eumelanin-to-pheomelanin (37.3%) of hair, and largest for total melanin (67%) of hair. Additionally, we refined our model-building algorithm, enabling visual comparison of the frequency and co-linearity due to linkage or co-inheritance of SNPs of the best-fitted models. Application of our algorithm to the test sample yielded the same or similar models as the training sample. Two of the three SNPs composing the models were the same, with some variability in the third SNP of the model.

Šiame darbe aš apibendrinau informaciją apie pasirinkimo tarp tiesinio prognozavimo mikro ir makro-modelių problemą. Agregavimas suprantamas kaip sektorinis agregavomas, o modeliai yra iš vienmatės tiesinės regresijos klasės. Aš išvedžiau kriterijų pasirinkimui tarp makro ir mikro-modelių ir idealaus agregavimo testą tiesinio agregavimo su fiksuotais ir atsitiktiniais svoriais atvejais. Paskutiniu atveju idealų agregavimą rekomenduoju tikrinti permutaciniu testu. Rezultatai iliustruoju ekonominiu pavyzdžiu. Modeliuoju Lietuvos vidutinį darbo užmokestį agreguotu modeliu ir atskirose ekonominės veiklos sektoriuose. Analizės rezultatas parodo, kad modeliai yra ekvivalentūs. / This paper focuses on the choice between macro and micro models. I suggest a hypothesis testing procedure for in-sample model selection for such variables as average wage. Empirical results show that Lithuanian average wage should be predict by using aggregate model.

The goal of this thesis is a financial analysis of the company ROLUX žaluzie, s.r.o.. The introductory part is dedicated to a detailed description of the basic methods and procedures of financial analysis based on company financial statements, presentation of prediction (bonity and bankruptcy) models and the Benchmarking diagnostic system of the Ministry of industry and trade of the Czech Republic - the INFA model. The practical-analytical part includes the application itself of the presented financial analysis procedures to the data of a real Czech company. Basic differences between the Czech accounting standard and IFRS that are relevant for the company being analyzed are also included in this work.

This thesis introduces the implementation of different supervised learning techniques for producing accurate estimates of soil moisture content using empirical information, including meteorological and remotely sensed data. The models thus developed can be extended to be used by the personal remote sensing systems developed in the Center for Self-Organizing Intelligent Systems (CSOIS). The dfferent models employed extend over a wide range of machine-learning techniques starting from basic linear regression models through models based on Bayesian framework. Also, ensembling methods such as bagging and boosting are implemented on all models for considerable improvements in accuracy. The main research objective is to understand, compare, and analyze the mathematical backgrounds underlying and results obtained from dfferent models and the respective improvisation techniques employed.

No description available.

Southern bottomland sites are among the most productive areas for producing high quality grade hardwood, yet the ability to estimate the quantity and quality of standing grade hardwood is almost non-existent. Measurements and observed log grades were recorded on standing trees to construct volume prediction models for individual trees. Several different modeling techniques were explored and compared during development. Developed equations predict merchantable sawtimber volume and volume by grade category in trees by species group. Two separate sets of equations were developed for each species group using either total height or merchantable height. Models were chosen based on significance of variables, index of fit, RMSE, bias, ease of use, and biological trends. The models developed to predict merchantable sawtimber and grade volumes were designed to be implemented in a larger hardwood growth and yield system.