Feature Selection is central to modern data science, from exploratory data analysis to predictive model-building. The "stability"of a feature selection algorithm refers to the robustness of its feature preferences, with respect to data sampling and to its stochastic nature. An algorithm is "unstable" if a small change in data leads to large changes in the chosen feature subset. Whilst the idea is simple, quantifying this has proven more challenging---we note numerous proposals in the literature, each with different motivation and justification. We present a rigorous statistical and axiomatic treatment for this issue. In particular, with this work we consolidate the literature and provide (1) a deeper understanding of existing work based on a small set of properties, and (2) a clearly justified statistical approach with several novel benefits. This approach serves to identify a stability measure obeying all desirable properties, and (for the first time in the literature) allowing confidence intervals and hypothesis tests on the stability of an approach, enabling rigorous comparison of feature selection algorithms.

The purpose of this study was to design an objective-based Anonymous Screening Process (ASP) for selecting the most qualified finalists for non-tenured administrative employment searches in public higher education. The intent was to design a viable employment process that substantially meets equal opportunity mandates, the dual aim of the Federal government's affirmative action efforts to eliminate the discriminatory effects of the past and to bar future discrimination, and the goal of the U.S. Congress of improving the economic status of disabled individuals by removing discriminatory barriers to full employment. An anonymous screening process is part of a comprehensive employment process, including recruitment, application, evaluation, and selection, wherein an applicant's name, race, religion, color, sex, age, handicap, and institutional identify are unknown to all persons involved in the process through selection of the most qualified finalists. An anonymous applicant screening process will minimize evaluator bias, effectively eliminate the interjection of non-objective criteria from external sources in selecting the most qualified finalists, and provide a defensible basis for using protected class status in the final selection as an additional criterion in situations of underrepresentation. / Graduation date: 1992

In personnel selection, employment tests are intended to reduce selection errors and increase mean performance. The current thesis examines the impact of psychometric properties of the predictors on selection accuracy, or the consistency between selection on observed scores versus true scores. Implications for validity and subsequent levels of job performance, or prediction accuracy, are also examined in light of common top-down personnel selection procedures. Results reflect the importance of having reliable and valid predictor measures; the work also extends ideas in the area of utility analysis.

South Africa has faced tremendous changes over the past decade, which has had a huge impact on the working environment. Organisations are compelled to address the societal disparities between various cultural groups. However, previously disadvantaged groups have had to face inequalities of the education system in the past, such as a lack of qualified teachers (especially in the natural sciences), and poor educational books and facilities. This has often resulted in poor grade 12 results. Social responsibility and social investment programmes are an attempt to rectify these inequalities. The objective of this research was to investigate the validity of the current selection battery of the Youth Foundation Training Programme (YFTP) in terms of academic performance of the students on the bridging programme. A correlational design was used in this research in order to investigate predictive validity whereby data on the assessment procedure was collected at about the time applicants were hired. The scores obtained from the Advanced Progressive Matrices (APM), which forms part of the Raven's Progressive Matrices as well as the indices of the Potential Index Battery (PIB) tests, acted as the independent variables, while the Matric results of the participants served as the criterion measure ofthe dependent variable. The data was analysed using the Statistical Package for Social Sciences (SPSS) software programme by means of correlations and regression analyses. The results showed that although the current selection battery used for the bridging students does indeed have some value, it only appears to be a poor predictor of the Matric results. Individually, the SpEEx tests used in the battery evidently were not good predictors of the Matric results, while the respective beta weights of the individual instruments did confirm that the APM was the strongest predictor. Limitations were identified and recommendations for further research were discussed. / Thesis (M.A. (Industrial Psychology))--North-West University, Potchefstroom Campus, 2007.

South Africa has faced tremendous changes over the past decade, which has had a huge impact on the working environment. Organisations are compelled to address the societal disparities between various cultural groups. However, previously disadvantaged groups have had to face inequalities of the education system in the past, such as a lack of qualified teachers (especially in the natural sciences), and poor educational books and facilities. This has often resulted in poor grade 12 results. Social responsibility and social investment programmes are an attempt to rectify these inequalities. The objective of this research was to investigate the validity of the current selection battery of the Youth Foundation Training Programme (YFTP) in terms of academic performance of the students on the bridging programme. A correlational design was used in this research in order to investigate predictive validity whereby data on the assessment procedure was collected at about the time applicants were hired. The scores obtained from the Advanced Progressive Matrices (APM), which forms part of the Raven's Progressive Matrices as well as the indices of the Potential Index Battery (PIB) tests, acted as the independent variables, while the Matric results of the participants served as the criterion measure ofthe dependent variable. The data was analysed using the Statistical Package for Social Sciences (SPSS) software programme by means of correlations and regression analyses. The results showed that although the current selection battery used for the bridging students does indeed have some value, it only appears to be a poor predictor of the Matric results. Individually, the SpEEx tests used in the battery evidently were not good predictors of the Matric results, while the respective beta weights of the individual instruments did confirm that the APM was the strongest predictor. Limitations were identified and recommendations for further research were discussed. / Thesis (M.A. (Industrial Psychology))--North-West University, Potchefstroom Campus, 2007.

Dissertation (Ph.D.)--Auburn University, / Abstract. Vita. Includes bibliographic references (leaves 44-55).

Diss. Wirtsch.-wiss. St. Gallen, 2000 ; Nr. 2464. / Literaturverz.

By definition "adaptability" is the ability of living systems to cope with change. Genetic adaptability requires the production of genetic variation. The view that variation production is undirected or random, i.e. unconnected with selection, implies that selection does not tailor genetic adaptability. But many genetic elements are known to modify processes of variation production, and secondary selection can act on them, so that view is not justified. Over the longer term, natural selection 'favors' properties important, in maintaining immediate fitness, as well as properties important for persistence in the short term. Genetic adaptability is less important in the short term, and is ignored in models based on short term definitions of fitness (e.g. relative effective rate of increase). If "fitness" is to be "the properties favored by natural selection", then its definition should be time scale dependent. Currently prevalent short term definitions of the action of natural selection should not be allowed to hamper consideration of the role of slow processes in determining long term evolutionary patterns. A review of patterns in genome size, and the existing explanations for them, reveals that most explanations are based on notions of adaptedness to the state of the environment. An explanation of genome size patterns based on the rate of change of environments is proposed. It is hypothesized that part of the genome is involved in regulating variation production, and that more DNA means slower production of additive genetic variation. This new hypothesis is simple, general, and testable, but requires more evidence. The question is raised of whether genomes might be organized to facilitate the adjustment of genetic variation production by natural selection. / Science, Faculty of / Zoology, Department of / Graduate

No description available.

Call number: LD2668 .T4 1928 R51