No description available.

No description available.

No description available.

This dissertation is the compilation of two major innovations that rely on a common technique known as multidimensional scaling (MDS). MDS is a dimension-reduction method that takes high-dimensional data and creates low-dimensional versions. Project 1: Visualizations are useful when learning from high-dimensional data. However, visualizations, just as any data summary, can be misleading when they do not incorporate measures of uncertainty; e.g., uncertainty from the data or the dimension reduction algorithm used to create the visual display. We incorporate uncertainty into visualizations created by a weighted version of MDS called WMDS. Uncertainty exists in these visualizations on the variable weights, the coordinates of the display, and the fit of WMDS. We quantify these uncertainties using Bayesian models in a method we call Informative Probabilistic WMDS (IP-WMDS). Visually, we display estimated uncertainty in the form of color and ellipses, and practically, these uncertainties reflect trust in WMDS. Our results show that these displays of uncertainty highlight different aspects of the visualization, which can help inform analysts. Project 2: Analysis of network data has emerged as an active research area in statistics. Much of the focus of ongoing research has been on static networks that represent a single snapshot or aggregated historical data unchanging over time. However, most networks result from temporally-evolving systems that exhibit intrinsic dynamic behavior. Monitoring such temporally-varying networks to detect anomalous changes has applications in both social and physical sciences. In this work, we simulate data from models that rely on MDS, and we perform an evaluation study of the use of summary statistics for anomaly detection by incorporating principles from statistical process monitoring. In contrast to most previous studies, we deliberately incorporate temporal auto-correlation in our study. Other considerations in our comprehensive assessment include types and duration of anomaly, model type, and sparsity in temporally-evolving networks. We conclude that the use of summary statistics can be valuable tools for network monitoring and often perform better than more involved techniques. / Doctor of Philosophy / In this work, two main ideas in data visualization and anomaly detection in dynamic networks are further explored. For both ideas, a connecting theme is extensions of a method called Multidimensional Scaling (MDS). MDS is a dimension-reduction method that takes high-dimensional data (all $p$ dimensions) and creates a low-dimensional projection of the data. That is, relationships in a dataset with presumably a large number of dimensions or variables can be summarized into a lower number of, e.g., two, dimensions. For a given data, an analyst could use a scatterplot to observe the relationship between 2 variables initially. Then, by coloring points, changing the size of the points, or using different shapes for the points, perhaps another 3 to 4 more variables (in total around 7 variables) may be shown in the scatterplot. An advantage of MDS (or any dimension-reduction technique) is that relationships among the data can be viewed easily in a scatterplot regardless of the number of variables in the data. The interpretation of any MDS plot is that observations that are close together are relatively more similar than observations that are farther apart, i.e., proximity in the scatterplot indicates relative similarity. In the first project, we use a weighted version of MDS called Weighted Multidimensional Scaling (WMDS) where weights, which indicate a sense of importance, are placed on the variables of the data. The problem with any WMDS plot is that inaccuracies of the method are not included in the plot. For example, is an observation that appears to be an outlier, really an outlier? An analyst cannot confirm this without further context. Thus, we created a model to calculate, visualize, and interpret such inaccuracy or uncertainty in WMDS plots. Such modeling efforts help analysts facilitate exploratory data analysis. In the second project, the theme of MDS is extended to an application with dynamic networks. Dynamic networks are multiple snapshots of pairwise interactions (represented as edges) among a set of nodes (observations). Over time, changes may appear in some of the snapshots. We aim to detect such changes using a process monitoring approach on dynamic networks. Statistical monitoring approaches determine thresholds for in-control or expected behavior that are calculated from data with no signal. Then, the in-control thresholds are used to monitor newly collected data. We applied this approach on dynamic network data, and we utilized a detailed simulation study to better understand the performance of such monitoring. For the simulation study, data are generated from dynamic network models that use MDS. We found that monitoring summary statistics of the network were quite effective on data generated from these models. Thus, simple tools may be used as a first step to anomaly detection in dynamic networks.

Whether the stated goal of a program is to improve health, reduce crime, or to increase standard of living, the ultimate goal of social programs is to improve overall quality of life. An adequate measure of this outcome would help determine whether achievement of these more specific goals (e.g., health, education) really leads to improvements in overall life quality, and would allow trade-offs to be made in terms of funding across programs. However, an understanding of the determinants of life quality (i.e., the mechanism by which a program did or did not have its intended effect) is also essential to program evaluation and the design of future programs.This study constitutes the analysis of an existing dataset of individual traits, life circumstances, satisfaction with a list of 30 life domains, and overall quality of life for 193 healthy elders to test a hypothesized model of the determinants of life quality. As expected, domain satisfaction appears to be a function of life circumstances. Individuals' traits (e.g., age, sex, personality) modify this relationship, but neither they, nor respondents' reports of domain importance, appear to have any direct effect on quality of life. Instead, domain satisfactions alone are the most proximal determinants of overall quality of life. It also appears that individuals respond differently in terms of overall quality of life to reductions in satisfaction with certain domains than to increases. These findings should be evaluated further as they could affect the design of future successful programs. Because individuals' traits and individuals' ratings of domain importance seem to have no effect on the relationship between domain satisfaction and overall quality of life, it may not be essential to measure these in future studies. Finally, although the data on life domains available to this study were sufficient to generate these results, the first step in the development of adequate measures of overall quality of life and of domain satisfactions will be the construction of a comprehensive, fully-representative list of the life domains that comprise life as a whole.

Policy-makers, project planners and development organisations are becoming convinced that the failure of the new socio-ecologically sensitive strategies co-opted by 'professional' forestry could be better addressed by indigenous forestry. They believe that indigenous forestry might assist with the development of successful forestry projects that are ecologically sustainable and socio-politically equitable. In order, however, to learn from indigenous forestry systems, the acculturation of foresters in the vernacular culture of the forest users appears to be an essential process for understanding and intervening in a local forest management complex. Acculturation entails not only more attention to the immaterial cultural realm, but an understanding of multiple resource stewardship, local ways of knowing and perceiving, local forest values and 'practices of care'. While acknowledging the significance of the politics of knowledge and political ecology this study examines resource stewardship from an alternative neglected angle that of knowledge sustainability and synergistic bridging. It will examine in general modes of knowing and bridging between 'formal' and indigenous forestry knowledge, and in particular the identification of forest value paradigms that are evidently exemplars of bio-cultural sustainability. The main outcomes of this study include the cognitive mapping of forest values among 'Tibetan minority nationalities' in Eastern Kham, their spatial distribution and the coincidence of changes in forest values with cultural or biophysical phenomena. Conceptually this study relies heavily on knowledge-system, hypertext, and paradigm theory and a critique of the narratives of John Locke. The former provide a platform to compare and contrast alternative knowledge systems and a means of synergistic bridging between them and the latter encapsulates a trajectory of western knowledge often known as modernity. The quantitative methods employed in this study included text analysis for forest value identification, multidimensional scaling for the cognitive mapping of forest values, spatial analysis and kriging for forest value distribution, and boundary or wombling analysis for changes in forest values and their coincidence with cultural or biophysical phenomena. The latter four methods are groundbreaking in that they have never been used to study forest values before. The study concludes that there is compelling evidence suggesting homogeneity in forest values with up to 5 geospatial paradigms and up to 12 cognitive paradigms. The findings, especially close correlation between forest values and ethnolinguistics, provide a potential template for foresters to develop multiple models of natural resource or biodiversity stewardship based on local forest values. In terms of the wider application, indigenous knowledge cannot seemingly be sustained if it is integrated with or into western knowledge systems due to the lack of conceptual frameworks for cross-cultural epistemological or psychological integration. Coalescing under the rubric of post-modernism, however, we do find a number of complimentary trajectories, which seemingly provide space for knowledge equity, sustainability and bridging. These trajectories include hypertext theory, paradigm theory, abductive logic, adaptive management, ecospiritual paradigms, and post-modern forestry paradigms. These trajectories and findings offer planners globally a means for synergistic bridging between local and non-local knowledge systems on the road to sustainable forestry and biodiversity stewardship.

Much of the research on metacognition in human factors has focused on prescriptive, normative strategy training. That is, many researchers have concentrated their efforts on finding ways to improve system users’ prediction, planning, monitoring and evaluation strategies for tasks. However little research has focused on the strategies and heuristics users employ on their own to make usability predictions. Understanding usability prediction methods is critical because users’ predictions inform their expectations about whether they will make errors using a product, how much effort they will need to expend to be successful in using the product, whether they can perform two tasks successfully at the same time, whether the costs of learning to use a device are worth the benefits of using it, which tools will assist in accomplishing goals and which tools will make performing the same task more difficult. The following study aims to identify the specific strategies people use to make usability predictions about product designs. From these strategies a set of guidelines, for designers who wish to ensure users’ expectations meet post hoc usability assessments, were proposed. The study was completed in two phases. During the first phase of this study, prediction strategies were elicited by 1) asking participants to make routine product usability judgments, from which implicit strategies can be inferred, and by 2) using explicit free-response methods. Judgments were analyzed using multi-dimensional scaling (MDS) methods to establish the number of dimensions that are implicitly used to predict usability. Subject matter experts (SMEs) coded free-response strategies using coding schemes developed in a pilot study. SMEs will also matched user strategies to formal, professional usability standards. The outcome of Phase 1 was usability taxonomy for classifying usability strategies that includes both expert and user language. The procedure was repeated with three different product design classes to determine how strategies differ as a function of the to-be-judged product. During the second phase of the study, a new group of participants rated specific usability attributes of designs to validate the strategies collected from users’ free-responses in Phase 1. Attributes were selected based on the strategies discovered in Phase 1. These usability attribute ratings helped to inform interpretations of the dimensions of the MDS model generated in Phase 1 and provided input into defining the usability attributes that influenced usability predictions. Results of this study reveal that the type design class participants evaluated had a significant effect on the type of strategy participants used to make their a priori usability assessments (UAs). Participants reported using “complexity” or “organization” most often to predict the usability of cookbooks. Participants reported using “mental simulation” or “typicality/familiarity” most often for predicting the usability of drinking glasses. Participants reported using “complexity,” “organization,” and to a lesser extent “typicality/familiarity,” and “mental simulation” as strategies for predicting the usability of cooktops. MDS methods were used to uncover the underlying dimension of the UA space. For drinking glasses, the “fanciness” and “holdability” were associated with UAs. For cooktops, “the number of controls” and whether participants believed “it was easy to understand how each burner was controlled” were associated with making UAs. And for cookbooks, “the length of the instructions” and “poor contrast of the text with the background” were associated with UAs. Overall, there is evidence that at least some participants in Phase 2 used terminology that was consistent with the terminology people used to describe the designs during Phase 1 and that these were congruent with the uncovered strategies.

Proximity scaling methods (e.g., multidimensional scaling) represent objects in a low dimensional configuration so that fitted distances between objects optimally approximate multivariate proximities. Next to finding the optimal configuration the goal is often also to assess groups of objects from the configuration. This can be difficult if the optimal configuration lacks clusteredness (coined c-clusteredness). We present Cluster Optimized Proximity Scaling (COPS), which attempts to solve this problem by finding a configuration that exhibts c-clusteredness. In COPS, a flexible scaling loss function (p-stress) is combined with an index that quantifies c-clusteredness in the solution, the OPTICS Cordillera. We present two variants of combining p-stress and Cordillera, one for finding the configuration directly and one for metaparameter selection for p-stress. The first variant is illustrated by scaling Californian counties with respect to climate change related natural hazards. We identify groups of counties with similar risk profiles and find that counties that are in high risk of drought are socially vulnerable. The second variant is illustrated by finding a clustered nonlinear representation of countries according to their history of banking crises from 1800 to 2010. (authors' abstract) / Series: Discussion Paper Series / Center for Empirical Research Methods

This thesis centers around dimensionality reduction and its usage on landmark-type data which are often used in anthropology and morphometrics. In particular we focus on non-linear dimensionality reduction methods - locally linear embedding and multidimensional scaling. We introduce a new approach to dimensionality reduction called multipass dimensionality reduction and show that improves the quality of classification as well as requiring less dimensions for successful classification than the traditional singlepass methods.

D.Litt. et Phil. / In recent times the South African society has been subject to rapid and important changes. These changes resulted in new responsibilities placed on social workers and psychologists. This situation has lead to the development of new perspectives and the expansion of knowledge and understanding. Social workers and psychologists are increasingly under pressure to provide cost effective services to an increasing number of clients without reducing accountability. Counsellors have to be able to support their decisions with scientific evidence ...