Epstein-Barr virus (EBV)-encoded latent membrane protein 1 (LMP1) plays important roles in EBV-mediated B cell transformation, development of EBV-associated malignancies, and exacerbation of certain autoimmune conditions. LMP1 functionally mimics tumor necrosis factor receptor (TNFR) superfamily member CD40, but LMP1 signals are amplified and sustained compared to those induced by CD40. CD40 and LMP1 rely on TNFR-associated factors (TRAFs) to mediate signaling, but use TRAFs differently. TRAF6 is important for CD40 signaling, and was implicated in LMP1 signaling in non-B cells. Here, we addressed the hypothesis that TRAF6 is a critical regulator of a subset of LMP1 signals in B cells. We found that TRAF6 was required for LMP1-mediated kinase activation and costimulatory molecule upregulation, and associated with the LMP1 TRAF1/2/3/5 binding site (TBS). Additionally, TRAF6 and the TBS contributed to LMP1-induced autoreactivity and antibody (Ab) production in vivo. Finally, in contrast to CD40, LMP1 required the TRAF6 receptor-binding domain to mediate TRAF6-dependent pathways. Thus, TRAF6 is critical for LMP1 signaling and requires LMP1 interaction to propagate signals. Importantly, TRAF6 associates with LMP1 in a manner distinct from CD40, raising the possibility of disrupting LMP1 functions while leaving normal CD40 signaling intact. We next investigated roles of the kinase TAK1 in TRAF6-dependent LMP1 functions. TAK1 was required for CD40- and LMP1-mediated JNK activation in B cells, leading to IL-6 and Ab production. Understanding mechanisms of CD40 and LMP1 signaling provides important insights into normal regulatory control of CD40 functions and how LMP1-mediated pathogenesis escapes or subverts these regulatory mechanisms. LMP1 itself may be a difficult therapeutic target, because it lacks an extracellular domain and is continually processed from the cell surface. Thus, it is important to elucidate similarities and differences between CD40 and LMP1 signals to identify therapeutic targets to block LMP1-mediated pathogenesis. Comparing and contrasting CD40 and LMP1 also increases our understanding of the critical mechanisms used to regulate normal CD40 signaling.

Latent Class Analysis was used to explore different subgroups of individuals based on identity processes (using the DIDS) and protestant fundamentalism. Results indicate that a 6-group solution provided the best fit for our data. The six groups differed in terms of identity process variables (especially exploration in breath, exploration in depth, and identification with commitment), but only modestly in terms of fundamentalism.

Spontaneous combustion of coal has historically been a major problem for the coal industry, predominantly during storage and transportation. Various methods have been used in the laboratory for evaluating the propensity of different coal sources to self-heat. However, the heterogeneity of coal and the complexity of the system has resulted in inconsistencies and sometimes conflicting results as indicated by the findings reported in several publications. The primary objective of the current study was to build a laboratory scale apparatus that simulates the condition of a coal stockpile to evaluate the events leading to spontaneous combustion and develop potential remedies. As such, the influential factors can be identified with confidence, thereby providing an improved understanding of the spontaneous combustion. An adiabatic heating apparatus was designed and constructed which included instrumentation to closely monitor the oxidation process and the stages leading to spontaneous combustion under various conditions. The device was equipped with thermocouples which measured the temperature rise as a function of time leading to the determination of an index value that indicated the propensity of a given coal source to spontaneously combustion. The index was referred to as the R70 value which was measured as the temperature was increased during the period of rapid oxidation. The units for the index was degrees Celsius per hour. As such, a high index value reflected the likelihood of spontaneous combustion for a given coal source. To standardize the test procedure, a detailed three-level statistical experimental design was conducted involving three critical parameters, i.e., particle size, oxygen flow rate and the duration of the drying period prior to feeding oxygen to the system. Using empirical models describing the R70 value as a function of the parameter values developed from the test data, it was determined that R70 was sensitive to the sample particle size and drying time. A decrease in particle size and drying time significantly increased the R70 value while the oxygen rate did not have a significant impact over the range of values tested. Based on the results of the test program, a standard test procedure was established to evaluate various coal sources and identify chemicals that could be used to remediate the spontaneous combustion issue. Several sub-bituminous coal sources collected from the Powder River Basin were tested in the apparatus and found to be prone to spontaneous combustion as indicated by R70 values that approached 50oC per hour. Several chemicals were evaluated as a means of eliminating or slowing the spontaneous combustion process. These agents included anti-oxidants, binders and humectants. Organic binders were used to agglomerate the fine coal particles which limited surface area exposure. The effect significantly reduced the oxidation rate as indicated by a reduction in the R70 index from 44.07oC/hr to 5.71oC/hr. However, after entering the latent heat stage, the temperature increased rapidly at a rate of 27.58oC/hr. Humectants were evaluated which contained several hydrophilic groups, mainly hydroxyl groups, and thus have an affinity for water. As a result, when the coals were treated with humectant, the latent heat rate was reduced to 4.24oC/hr although the R70 remained relatively high. By using a combination of humectant and binder, the optimum result was obtained with an R70 value of 5.04oC/hr and a latent heat rate of 11.06oC/hr. These findings were successfully implemented into industrial practice for significantly delaying the spontaneous combustion event.

While a growing evidence base suggests that expressive writing about a traumatic event may be an effective intervention which results in a variety of health benefits, there are still multiple competing theories that seek to explain expressive writing’s mechanism(s) of action. Two of the theories with stronger evidence bases are exposure theory and cognitive processing theory. The state of this field is complicated by methodological limitations; operationalizing and measuring the relative constructs of trauma narratives, such as coherence, traditionally requires time- and labor-intensive methods such as using a narrative coding scheme. This study used a computer-based methodology, latent semantic analysis (LSA), to quantify narrative coherence and analyze the relationship between narrative coherence and both short- and long-term outcomes of expressive writing. A subsample of unscreened undergraduates (N=113) who had been randomly assigned to the expressive writing group of a larger study wrote about the most traumatic event that had happened to them for three twenty-minute sessions; their narratives were analyzed using LSA. There were three main hypotheses, informed by cognitive processing theory: 1) That higher coherence in a given session would be associated with a more positive reported valence at the conclusion of that session, 2) that increasing narrative coherence across writing sessions would be associated with increasing reported valence at the conclusion of each session, and 3) that increasing narrative coherence over time would be associated with a decrease in post-traumatic stress symptoms. Overall, initial hypotheses were not supported, but higher coherence in the third writing session was associated with more negative valence at the conclusion of the session. Furthermore, relationships between pre- and post-session valence strengthened over time, and coherence, pre-session valence, and post-session valence all trended over time. These results suggest a collection of temporal effects, the implications of which are discussed in terms of future directions.

Writing self-efficacy is a vital component to a students’ motivation and will to succeed towards writing. The measurement of writing self-efficacy over the past 40 years, despite its development, continues to largely be represented by Confirmatory Factor Analysis models that are limited due to their restricted item to factor constraints. These constraints, given prior literature and the theoretical understanding of self-efficacy, do not adequately model construct- relevant psychometric multidimensionality as a product of conceptual overlap or a hierarchical or general factor. Given this, the present study’s purpose was to examine the adapted Self-efficacy for Writing Scale (SEWS) for the presence of construct-relevant psychometric multidimensionality through a series of measurement model comparisons and person-centered approaches. Using a sample 1,466 8th, 9th, and 10th graders, a bifactor exploratory structural equation model was found to best represent the data and demonstrate that the SEWS exhibits both construct-relevant multidimensionality as a function of conceptual overlap and the presence of a hierarchical theme. Using factor scores derived from this model, latent profile analysis was conducted to further establish validity of the measurement model and examine how students disaggregate into groups based on their response trends of the SEWS. Three profiles emerged greatly differentiated by global writing self-efficacy, with obvious and substantively varying specific factor differences between profiles. Concurrent, divergent, and discriminant validity evidence was established through a series of analyses that assessed predictors and outcomes of the profiles (e.g. demographics, standardized writing assessments, grades). Theoretical and educator implications and avenues for future researcher were discussed.

The purpose of this dissertation was to expand the understanding of youth in psychiatric residential treatment facilities by using psychosocial indicators to develop subgroup profiles. Additionally, differences in treatment outcomes between subgroups and the extent to which within-treatment factors accounted for observed differences in treatment outcomes between subgroups were examined. Data were extracted from the case records of 447 youth who were served in psychiatric residential treatment facilities over a seven year span of time. A latent class analysis was used to identify and describe subgroups. A series of multivariate regression analyses were used to examine group differences in functional impairment at discharge. Next, a path analysis was used to determine if there were differences in average change in functional impairment from admission to discharge between subgroups and to test within treatment factors as potential mediators of group differences. Finally, a logistic regression was used to determine if there were differences between groups in the probability of discharging to a community-based placement or discharging to another congregate care facility. The latent class analysis revealed four distinct subgroups of youth. The analyses of treatment outcomes revealed statistically significant differences in the level of functional impairment at discharge and average change in impairment between groups. Results from the path model of indirect effects supported that within treatment factors accounted for a statistically significant proportion of the observed difference in change between groups. No differences were found in discharge placement outcomes between groups. Implications for future research, practice and policies focused on youth in residential treatment are discussed.

In this dissertation, I explored relational learning via latent variable models. Traditional machine learning algorithms cannot handle many learning problems where there is a need for modeling both relations and noise. Statistical relational learning approaches emerged to handle these applications by incorporating both relations and uncertainties in these problems. Latent variable models are one of the successful approaches for statistical relational learning. These models assume a latent variable for each entity and then the probability distribution over relationships between entities is modeled via a function over latent variables. One important example of relational learning via latent variables is text data modeling. In text data modeling, we are interested in modeling the relationship between words and documents. Latent variable models learn this data by assuming a latent variable for each word and document. The co-occurrence value is defined as a function of these random variables. For modeling co-occurrence data in general (and text data in particular), we proposed latent logistic allocation (LLA). LLA outperforms the-state-of-the-art model --- latent Dirichlet allocation --- in text data modeling, document categorization and information retrieval. We also proposed query-based visualization which embeds documents relevant to a query in a 2-dimensional space. Additionally, I used latent variable models for other single-relational problems such as collaborative filtering and educational data mining. To move towards multi-relational learning via latent variable models, we propose latent feature networks (LFN). Multi-relational learning approaches model multiple relationships simultaneously. LFN assumes a component for each relationship. Each component is a latent variable model where a latent variable is defined for each entity and the relationship is a function of latent variables. However, if an entity participates in more than one relationship, then it will have a separate random variable for each relationship. We used LFN for modeling two different problems: microarray classification and social network analysis with a side network. In the first application, LFN outperforms support vector machines --- the best propositional model for that application. In the second application, using the side information via LFN can drastically improve the link prediction task in a social network.

Developmental and prevention researchers aim to determine how unhealthy behaviors emerge. Mediation analysis offers a statistical tool that allows researchers to describe the processes underlying early risk and later health outcomes. Among existing longitudinal mediation models, latent difference score mediation stands out due to its unique ability to capture variations in changes both within and across individuals, as well as its ability to examine non-linear change over time. However, the literature currently lacks sample size guidelines for latent difference mediation models, which has proven to make the use of these models difficult. The current project addresses this limitation by offering an empirical set of sample guidelines for a variety of latent difference mediation score models through a Monte Carlo simulation study. By offering empirical sample size guidelines for latent difference score mediation models, future developmental and prevention researchers can make informed sampling decisions prior to data collection. Moreover, women who misuse alcohol have been found to experience more severe medical consequences than men. However, minimal research has evaluated how gender specific risk factors influence its onset. The current project addresses this limitation by applying latent difference score mediation to evaluate how disordered eating behaviors among adolescent girls influence alcohol misuse among adult women.

Monitoring the dynamic thermal activity in Yellowstone National Park is required by the United States Congress. The continuous monitoring is important to maintain the safety of the visitors and park service personnel, plan and relocate infrastructure, and study potential impact from nearby geothermal development including oil and gas industry. This dissertation is part of a study initiated in the early 2000s to monitor the thermal activity of dynamic areas within the Park, using airborne remote sensing imagery. This study was focused in Norris Geyser Basin, the hottest geyser basin in the park, located near the northwestern rim of the Yellowstone’s caldera. The study is considered the first long-term comprehensive airborne remote sensing study in the basin which took place between August 2008 and October 2013. In this study, at least one 1-meter resolution thermal infrared image and three-band images (multispectral) were acquired and used to estimate year-to-year changes in radiant temperature, radiant flux, and radiant power from the thermal source in Norris. Presence of residual radiant flux in the ground from absorbed solar radiation and atmospheric longwave radiation was the main challenge to compere year-to-year changes in the thermal activity. This residual flux is included in the total radiant flux calculated through the remote sensing images which gives false estimates of the flux generated from the underling thermal source. Two methods were suggested in Chapters 2 and 4 of this dissertation to estimate the residual radiant flux. A method was developed in Chapter 2 to estimate the residual radiant flux in a bare ground area covered with hydrothermal siliceous sinter deposit. The method compared ground-based measurements with high spatial resolution airborne remote sensing measurements to estimate the residual radiant flux. In Chapter 4, a method was developed to estimate the residual radiant flux in the six surface classes in Norris, including bare ground, bare ground with siliceous sinter deposit, lakes and pools, river, forest, and grass. The assumptions and implications of each method were discussed to suggest a reliable method to estimate the geothermal radiant flux after subtracting the absorbed residual radiant flux. Chapter 3 provides an analysis of the four components of heat flux in the ground surface, including conduction of sensible heat, convection of sensible heat by liquid water and water vapor, and convection of latent heat by water vapor. The main purpose from the analysis was to assess the hypothesis that the convection and latent heat flux are negligible which therefore supported the results obtained from the analysis in Chapters 2 and 4.

Factor mixture modeling (FMM) has been increasingly used to investigate unobserved population heterogeneity. This Monte Carlo simulation study examined the issue of measurement invariance testing with FMM when there are covariate effects. Specifically, this study investigated the impact of excluding and misspecifying covariate effects on the class enumeration and measurement invariance testing with FMM. Data were generated based on three FMM models where the covariate had impact on the latent class membership only (population model 1), both the latent class membership and the factor (population model 2), and the latent class membership, the factor, and one item (population model 3). The number of latent classes was fixed at two. These two latent classes were distinguished by factor mean difference for conditions where measurement invariance held in the population, and by both factor mean difference and intercept differences across classes when measurement invariance did not hold in the population. For each of the population models, different analysis models that excluded or misspecified covariate effects were fitted to data. Analyses consisted of two parts. First, for each analysis model, class enumeration rates were examined by comparing the fit of seven solutions: 1-class, 2-class configural, metric, and scalar, and 3-class configural, metric, and scalar. Second, assuming the correct solution was selected, the fit of analysis models with the same solution was compared to identify a best-fitting model. Results showed that completely excluding the covariate from the model (i.e., the unconditional model) would lead to under-extraction of latent classes, except when the class separation was large. Therefore, it is recommended to include covariate in FMM when the focus is to identify the number of latent classes and the level of invariance. Specifically, the covariate effect on the latent class membership can be included if there is no priori hypothesis regarding whether measurement invariance might hold or not. Then fit of this model can be compared with other models that included covariate effects in different ways but with the same number of latent classes and the same level of invariance to identify a best-fitting model.