Thesis (M.A.)--Trinity International University, 2000. / Abstract. Includes bibliographical references (leaves 141-147).

Thesis (M.A.)--Trinity International University, 2000. / Abstract. Includes bibliographical references (leaves 141-147).

The present study extends previous research on self-control procedures and the transformation of stimulus functions. Using a multiple baseline design, participants were exposed to a relational responding task in which attempted to manipulate the functions of stimuli. They were exposed to a preference assessment, a naturalistic baseline, and a choice baseline before the relational training and testing were implemented. This procedure established contextual cues of more-than and less-than when paired with multiple exemplars of the stimuli during training. Re-exposure to the choice baseline was introduced in order to support the transformation of functions. Results are discussed regarding the alteration of preferences from the small, immediate reinforcer initially, to the larger, more delayed reinforcer following the intervention. Implications of the study and future research are also discussed.

Alcohol dependence is a complex disease involving polygenes, environment and their interactions. Inadequate consideration of these interactions may have hampered the progress on genome-wide association studies of alcohol dependence. By using the dataset of the Study of Addiction: Genetics and Environment with 3838 subjects, we conducted a genome-wide association studies of alcohol dependence symptom count (ADSC) with a full genetic model considering additive, dominance, epistasis and their interactions with ethnicity, as well as conditions of co-morbid substance dependence. Twenty quantitative trait single nucleotide polymorphisms (QTSs) showed highly significant associations with ADSC, including four previously reported genes (ADH1C, PKNOX2, CPE and KCNB2) and the reported intergenic rs1363605, supporting the overall validity of the analysis. Two QTSs within or near ADH1C showed very strong association in a dominance inheritance mode and increased the phenotype value of ADSC when the effect of co-morbid opiate or marijuana dependence was controlled. Highly significant association was also identified in variants within four novel genes (RGS6, FMN1, NRM and BPTF), two non-coding RNA and two epistasis loci. QTS rs7616413, located near PTPRG encoding a protein tyrosine phosphatase receptor, interacted with rs10090742 within ANGPT1 encoding a protein tyrosine phosphatase in an additive × additive or dominance × additive manner. The detected QTSs contributed to about 20 percent of total heritability, in which dominance and epistasis effects accounted for over 50 percent. These results demonstrated that perturbations arising from gene–gene interaction and conditions of co-morbidity substantially influence the genetic architecture of complex trait.

Conditional coding is an application of Markov Chain Monte Carlo methods for sampling from conditional distributions. It is applied here to the problem of estimating the parameters of a computer-simulated pattern of fractures in an isomorphic, homotropic material under plane strain. We investigate the theory and procedures of conditional coding and show the viability of the technique by its application.

Reasoning from all knowledge and belief is an adaptive approach to thinking about the world. It has been robustly shown that conditional ‘if then’ reasoning with everyday content is influenced by the background knowledge an individual has available. If we are presented with the statement ‘if it rains, then John will get wet’ then we are told that it is raining and asked if John will get wet, we may consider a number of possibilities before answering the question; perhaps John has an umbrella or is sheltered from the rain. Hence, when engaged in conditional reasoning of this sort people typically draw on background knowledge to arrive at an informed response. People with autism tend not to process information in context. There is a wealth of evidence indicating that these individuals have a piecemeal rather than an integrative processing style. It was therefore hypothesised that adolescents with autism spectrum disorder (ASD) would be less influenced by background knowledge when engaged in conditional reasoning with everyday content. Adolescents with ASD showed a weak or absent effect of available background knowledge on reasoning outcomes compared to a typically developing control group. This finding was demonstrated in two separate conditional reasoning tasks. These results were not explained by a failure to generate background knowledge or by differences in the beliefs held by the two groups regarding problem content. Within the typical population a lack of contextualised reasoning was also found among participants with high scores on one particular autistic trait, attention to detail. The ability to integrate all relevant information during conditional reasoning was also found to be dependent on available working memory resources. These results extend the known domains which demonstrate a lack of contextualised processing in autism. They also show that for individuals with autism reasoning without regard for background knowledge stems from a failure to integrate information. The findings suggest that this failure is related to the cognitive demands of the task and the processing style of the individual.

This thesis encompasses three essays around criminal behaviour with the first one analysing the impact of programmes aimed at poverty reduction, the second one developing a theoretical model of hysteresis in crime, and the third one empirically investigating the hysteresis hypothesis in crime rates. In the first chapter I investigate the impact of conditional cash transfers (CCT) on crime rates by analysing the Brazilian Bolsa Familia, the largest CCT programme in the world, in a panel data between 2001 and 2008. The related existing economic literature analysing general welfare programmes usually ignores the crucial endogeneity involved in the relationship between crime rates and social welfare policies through poverty, since poorer regions are focused in the distribution of resources. I use the existing temporal heterogeneity in the implementation of the programme across the states to identify the causal impact of CCT programmes on poverty and criminality. The guidelines of the Brazilian programme established that the amount of resources available for each state should be based on the poverty levels in the 2000 Census. However, due to reasons unrelated to poverty levels and crime rates, some states were able to implement the programme to a greater extent more quickly than others. States that reached the level of cash transfer expenditures proposed by the guidelines of the programme more promptly had a more significant reduction in poverty rates. Similar but less robust results are found for crime rates as robbery, theft and kidnapping, while no significant effects were found for homicide and murder, indicating a weak or non-existent relationship between conditional cash transfers and crime. I also develop, to my knowledge, the first theoretical model to explicitly account for hysteresis - a situation where positive exogenous variations in the relevant economic variables have a different effect from negative variations - in both criminal behaviour and crime rates in order to fill the gap between the theoretical predictions and the empirical evidence about the efficiency of policies in reducing crime rates. The majority of the theoretical analyses predict a sharp decrease in crime rates when there are significant improvements in the economic conditions or an increase in the probability of punishment. However, the existing empirical studies have found lower than expected effects on crime rates from variations in variables related to those factors. One important consequence of hysteresis is that the effect on an outcome variable from positive exogenous variations in the determining variables has a different magnitude from negative variations. For example, if hysteresis is present in the criminal behaviour and part of the police force in a city are dismissed in a given year, resulting in an escalation in crime, a reversal of the policy in the following year by readmitting all sacked police officers in an attempt to restore the original crime levels will result in lower crime rates, but higher than the original ones, yielding an asymmetric relationship between police and crime. Hysteresis is considered in a simple framework to model illicit behaviour. At the individual level, if criminal activity is associated with intrinsic sunk costs and learning, then the cost of leaving a criminal career is higher than entering it. At the aggregate level with homogeneous agents, this is translated into a hysteresis effect that will only occur if a specific threshold is surpassed. With heterogeneous agents, this phenomenon is reinforced generating a hysteresis effect that exists for all possible values of the variable affecting the crime decision. There are multiple equilibria at both levels. In the last chapter I empirically investigate the existence of hysteresis in crime rates. To my knowledge, this is the first empirical study to consider the existence of asymmetric effects on crime from variations in the probability of punishment and in the opportunity cost of crime. More specifically, I investigate whether positive variations on variables associated to those factors, respectively police officers and average level of income, are statistically different from negative variations. Using US crime data at the state level between 1977 and 2010, I find that police force size and real average income of unskilled workers have asymmetric effects on most types of crimes. The absolute value of the average impact of positive variations in those variables on property and violent crime rates are statistically smaller than the absolute value of the average effect of negative variations. These effects are robust under several specifications. A closer inspection of the data reveals a relatively monotonic negative relationship between wages and property crime rates, as well as negative variations in police and most crime rates. However, the relationships between positive variations in law enforcement size and most crime rates are non-linear. The magnitude of the observed asymmetries supports the hypothesis of hysteresis in crime, and suggests that no theoretical or empirical analysis would be complete without careful consideration of that important feature in the relationships between crime, police and legal income. These results corroborate the argument that policy makers should be more inclined to set pre-emptive policies rather than mitigating measures.

Bayesian subset selection suffers from three important difficulties: assigning priors over model space, assigning priors to all components of the regression coefficients vector given a specific model and Bayesian computational efficiency (Chen et al., 1999). These difficulties become more challenging in Bayesian quantile regression framework when one is interested in assigning priors that depend on different quantile levels. The objective of Bayesian quantile regression (BQR), which is a newly proposed tool, is to deal with unknown parameters and model uncertainty in quantile regression (QR). However, Bayesian subset selection in quantile regression models is usually a difficult issue due to the computational challenges and nonavailability of conjugate prior distributions that are dependent on the quantile level. These challenges are rarely addressed via either penalised likelihood function or stochastic search variable selection (SSVS). These methods typically use symmetric prior distributions for regression coefficients, such as the Gaussian and Laplace, which may be suitable for median regression. However, an extreme quantile regression should have different regression coefficients from the median regression, and thus the priors for quantile regression coefficients should depend on quantiles. This thesis focuses on three challenges: assigning standard quantile dependent prior distributions for the regression coefficients, assigning suitable quantile dependent priors over model space and achieving computational efficiency. The first of these challenges is studied in Chapter 2 in which a quantile dependent prior elicitation scheme is developed. In particular, an extension of the Zellners prior which allows for a conditional conjugate prior and quantile dependent prior on Bayesian quantile regression is proposed. The prior is generalised in Chapter 3 by introducing a ridge parameter to address important challenges that may arise in some applications, such as multicollinearity and overfitting problems. The proposed prior is also used in Chapter 4 for subset selection of the fixed and random coefficients in a linear mixedeffects QR model. In Chapter 5 we specify normal-exponential prior distributions for the regression coefficients which can provide adaptive shrinkage and represent an alternative model to the Bayesian Lasso quantile regression model. For the second challenge, we assign a quantile dependent prior over model space in Chapter 2. The prior is based on the percentage bend correlation which depends on the quantile level. This prior is novel and is used in Bayesian regression for the first time. For the third challenge of computational efficiency, Gibbs samplers are derived and setup to facilitate the computation of the proposed methods. In addition to the three major aforementioned challenges this thesis also addresses other important issues such as the regularisation in quantile regression and selecting both random and fixed effects in mixed quantile regression models.

Limitations imposed on control functions can significantly affect the performance of a linear controller. When applied to the real physical system, such limitations convert a linear function to a nonlinear input signal that alters the convergence or stability of the solution. The main focus of this study is to identify, classify and propose appropriate techniques to overcome such problems. In this regard, we propose an exact definition for a conditionally controllable problem and investigate control function formulations for such problems under the lenses of planning-based and optimization-based methods.

Dans cette thèse nous étudions l'estimation de modèles probabilistes discriminants, surtout des aspects d'apprentissage semi-supervisé et de sélection de caractéristiques. Le but de l'apprentissage semi-supervisé est d'améliorer l'efficacité de l'apprentissage supervisé en utilisant des données non-étiquetées. Cet objectif est difficile à atteindre dans les cas des modèles discriminants. Les modèles probabilistes discriminants permettent de manipuler des représentations linguistiques riches, sous la forme de vecteurs de caractéristiques de très grande taille. Travailler en grande dimension pose des problèmes, en particulier computationnels, qui sont exacerbés dans le cadre de modèles de séquences tels que les champs aléatoires conditionnels (CRF). Notre contribution est double. Nous introduisons une méthode originale et simple pour intégrer des données non étiquetées dans une fonction objectif semi-supervisée. Nous démontrons alors que l'estimateur semi-supervisé correspondant est asymptotiquement optimal. Le cas de la régression logistique est illustré par des résultats d'expèriences. Dans cette étude, nous proposons un algorithme d'estimation pour les CRF qui réalise une sélection de modèle, par le truchement d'une pénalisation $L_1$. Nous présentons également les résultats d'expériences menées sur des tâches de traitement des langues (le chunking et la détection des entités nommées), en analysant les performances en généralisation et les caractéristiques sélectionnées. Nous proposons finalement diverses pistes pour améliorer l'efficacité computationelle de cette technique.