Browsing by Subject "Statistics"
Now showing 1 - 6 of 6
Results Per Page
Sort Options
Item EFFECTS OF "HIGH-IMPACT PRACTICES" ON FIRST-GENERATION COLLEGE STUDENTS' ACADEMIC SUCCESS(2019-01-01) Gregg, Delana S; Bickel, Beverly; Arnold Lincove, Jane; Language, Literacy & Culture; Language Literacy and CultureDemographic changes in the U.S. indicate that more first-generation college students will matriculate into higher education in the next decade. First-generation college students may face a number of social and cultural barriers to success in college, and are less likely to graduate nationally. "High-impact practices" have been proffered as interventions that can help students, especially first-generation college students, build academic, social and psychological engagement in college and help them succeed. This quantitative study estimates the effects for first-generation college students from participation in five specific and widely used "high-impact practices": academic first-year seminars, extended orientation first-year seminars, service learning, internships and living learning communities. Utilizing propensity score matching on student level variables, this research project more accurately estimates treatment effects of "high-impact practices," controlling for endogeneity based on self-selection, bias often unaccounted for in much of the extant quantitative research on "high-impact practices." Five years of longitudinal student data, including 15,828 student records from 2013-2018, were analyzed from UMBC, a mid-sized public research university. When combined with 9,985 survey responses from students who participated in "high-impact practices" (at the same university during the same years), this study helps explain how such interventions may help students succeed. The research reveals that for first-time first-generation college students, service learning and internships have statistically significant effects on student success (final grade point average, persistence and graduation). For transfer first-generation college students, extended-orientation first-year seminars, service learning and internships were associated with success. Internships were associated with larger positive treatment effects for first-generation college students. Combined with the survey study of students' perceptions of the effects of participating in "high-impact practices," these same three practices which were statistically significantly associated with student success also were reported by students to help them to feel engaged, be motivated to graduate, and build critical thinking and interpersonal communication skills, including teamwork and networking skills. This research contributes to the scholarship on "high-impact practices," providing more accurate measures of treatment effects on first-generation college students' academic success and contributing to the understanding of how specific "high-impact practices" affect different students' academic, social and psychological engagement in college.Item Federal Government Document Statistics for 2014-15Hartman, LisaItem Mapping the structure of knowledge for teaching nominal categorical data analysis(2013) Groth, Randall E.; Bergner, Jennifer A.This report describes a model for mapping cognitive structures related to content knowledge for teaching. The model consists of knowledge elements pertinent to teaching a content domain, the nature of the connections among them, and a means for representing the elements and connections visually. The model is illustrated through empirical data generated as prospective teachers were in the process of developing knowledge for teaching nominal categorical data analysis. During a course focused on the development of statistical knowledge for teaching, the prospective teachers analyzed statistical problems, descriptions of children’s statistical thinking, and related classroom scenarios. Their analyses suggested various types of knowledge structures in development. In some cases, they constructed all knowledge elements targeted in the course. In many cases, however, their knowledge structures had missing, incompatible, and/or disconnected elements preventing them from carrying out recommendations for teaching elementary nominal categorical data analysis in an optimal manner. The report contributes to teacher education by drawing attention to prospective teachers’ learning needs, and it contributes to research on teachers’ cognition by providing a method for modeling their cognitive structures.Item Modeling Overdispersion in R(2015) Raim, Andrew M.; Neerchal, Nagaraj K.; Morel, Jorge G.The book Overdispersion Models in SAS by Morel and Neerchal (2012) discusses statistical analysis of categorical and count data which exhibit overdispersion, with a focus on computational procedures using SAS. This document retraces some of the ground covered in the book, which we abbreviate throughout as OMSAS, with the objective of carrying out similar analyses in R (R Core Team, 2014). Rather than attempting to cover every example in OMSAS, we will focus on two specific goals: analysis based on binomial/multinomial likelihoods which support extra variation, and model selection with the binomial goodness-of-fit (GOF) test. We will not cover examples based on count data, but extension to those should not be difficult. We will generally not spend much time discussing the data, on justification for the selected models, or on interpretation of the results. The reader should refer to OMSAS for more complete discussions of the examples and statistical models. In several places we will present additional material not found in OMSAS, such as the binomial finite mixture and the recently proposed Mixture Link binomial model.Item Summary Verification Measures and Their Interpretation for Ensemble Forecasts(American Meteorological Society, 2011-09-01) Bradley, A. Allen; Schwartz, Stuart S.Ensemble prediction systems produce forecasts that represent the probability distribution of a continuous forecast variable. Most often, the verification problem is simplified by transforming the ensemble forecast into probability forecasts for discrete events, where the events are defined by one or more threshold values. Then, skill is evaluated using the mean-square error (MSE; i.e., Brier) skill score for binary events, or the ranked probability skill score (RPSS) for multicategory events. A framework is introduced that generalizes this approach, by describing the forecast quality of ensemble forecasts as a continuous function of the threshold value. Viewing ensemble forecast quality this way leads to the interpretation of the RPSS and the continuous ranked probability skill score (CRPSS) as measures of the weighted-average skill over the threshold values. It also motivates additional measures, derived to summarize other features of a continuous forecast quality function, which can be interpreted as descriptions of the function’s geometric shape. The measures can be computed not only for skill, but also for skill score decompositions, which characterize the resolution, reliability, discrimination, and other aspects of forecast quality. Collectively, they provide convenient metrics for comparing the performance of an ensemble prediction system at different locations, lead times, or issuance times, or for comparing alternative forecasting systems.Item UNIVARIATE CLASSIFICATION OF DEGRADED CARTILAGE SAMPLES USING MRI PARAMETERS(2012-01-01) GURLU, MERVE; Sinha, Bimal K.; Spencer, Richard; Mathematics and Statistics; StatisticsEarly diagnosis of osteoarthritis is important in implementing treatments to control the progression of the disease. In this thesis, we applied univariate classification methods to distinguish healthy and degraded cartilage samples using five MRI parameters. We first replicated the univariate classification method used in the industry. We then tried to improve specificity and sensitivity by introducing new classification methods: standardized distance, likelihood ratio test, and likelihood ratio test with noise added to the validation set. We applied the conventional and the new classification methods to control, 18-h trypsin-digested, and 20-h collagenase-digested samples. Two separate analyses were performed; control group combined with trypsin-digested group and the same control group combined with collagenase-digested group. Specificity and sensitivity were computed and the results of all four methods were compared. Furthermore, a simulation analysis was done to compare the accuracy rate of the conventional method and the likelihood ratio test.