Search

At Wake Forest University, a student who is blind enrolled in a second course in statistics. The course covered simple and multiple regression, model diagnostics, model selection, data visualization, and elementary logistic regression. These topics required that the student both interpret and produce three sets of materials: mathematical writing, computer programming, and visual displays of data. While we did find scattered resources for blind students taking mathematics courses or introductory statistics courses, we found no complete account of teaching statistical modeling to students who are blind. We also discovered some challenges in stitching together multiple partial solutions. This paper outlines our specific approach. We relied heavily on integrating the use of multiple existing technologies. Specifically, this paper will detail the extensive use of screen readers, LATEX, a modified use of R and the BrailleR package, a desktop Braille embosser, and a modified classroom approach.

In 2009, the North Carolina Community College System (NCCCS) began the process of redesigning developmental mathematics (DMA) courses. The first classes were offered in 2012 with final state implementation in fall 2013. The purpose of this study was to examine the subsequent gateway course success rates achieved by former DMA students in relation to the delivery method (i.e., teacher-centered, student-centered, and computer-centered) used in the redesigned North Carolina DMA courses. Students included in this study took a gateway math course and a developmental math course during the same or previous academic year. Students who participated in student-centered and computer-centered instruction in their DMA course had a statistically significant pass rate in subsequent gateway courses compared with students receiving teacher-centered instruction. Along with the findings, this paper discusses considerations for redesign, including the choice of delivery methods and instructional techniques, and recommendations for further research.

In a recidivism prediction context, there is no consensus on which modeling strategy should be followed for obtaining an optimal prediction model. In previous papers, a range of statistical and machine learning techniques were benchmarked on recidivism data with a binary outcome. However, two important tree ensemble methods, namely gradient boosting and random forests were not extensively evaluated. In this paper, we further explore the modeling potential of these techniques in the binary outcome criminal prediction context. Additionally, we explore the predictive potential of classical statistical and machine learning methods for censored time-to-event data. A range of statistical manually specified statistical and (semi-)automatic machine learning models is fitted on Dutch recidivism data, both for the binary outcome case and censored outcome case. To enhance generalizability of results, the same models are applied to two historical American data sets, the North Carolina prison data. For all datasets, (semi-) automatic modeling in the binary case seems to provide no improvement over an appropriately manually specified traditional statistical model. There is however evidence of slightly improved performance of gradient boosting in survival data. Results on the reconviction data from two sources suggest that both statistical and machine learning should be tried out for obtaining an optimal model. Even if a flexible black-box model does not improve upon the predictions of a manually specified model, it can serve as a test whether important interactions are missing or other misspecification of the model are present and can thus provide more security in the modeling process. [ABSTRACT FROM AUTHOR]

National reports have documented deficiencies in the vertical alignment of mathematical learning in K-12 education. Many students fail to master requisite concepts before advancing to more complex ideas, leaving them ill-prepared to succeed in higher level Science, Technology, Engineering, and Mathematics (STEM) coursework. In this paper, we model elementary and middle school students’ performance in mathematics over time as a stochastic process to forecast their proficiency by the eighth grade. We conduct an extensive examination of tens of thousands of student records and extract useful information. We use this data to present a longitudinal analysis of student performance on the North Carolina End-of-Grade mathematics exam and use Markov chain models to probabilistically characterize the movement of students’ scores from one grade level to the next. This work is the first step in developing a framework to forecast individual students’ development of mathematical knowledge over time. [ABSTRACT FROM AUTHOR]

Background: Concurrent sexual partnerships (partnerships that overlap in time) may contribute to higher rates of HIV transmission in African Americans. Attitudes toward a behavior constitute an important component of most models of health-related behavior and behavioral change. We have developed a scale, employing realistic vignettes that appear to reliably measure attitudes about concurrency in young African American adults. Methods: Vignette-based items to assess attitudes about concurrency were developed following focus groups and cognitive testing of items adapted from existing scales assessing psychosocial constructs surrounding related sexual behaviors. The new items were included in a telephone survey of African American adults (18–34 years old) in Eastern North Carolina immediately before and after a radio campaign designed to discourage concurrency. We performed an exploratory factor analysis on each sample (pre- and post-campaign) to cross-validate results. We retained factors with a primary loading of ≥0.50 and no secondary loading >0.30. Cronbach’s coefficient alpha was used to evaluate internal reliability. Associations in the predicted direction between the mean responses to items on the final factor and known correlates of concurrency validated the scale. Results: Factor analysis in a random pre-campaign subsample yielded a one-factor 6-item scale with acceptable internal consistency (Cronbach’s α = 0.79). As expected, the attitude factor was positively associated with participation in concurrent partnerships, whether assessed by self-report (r = 0.298, p<0.0001) or deduced from dates of recent sexual partnerships (r = 0.298, p<0.0001). The factor was also positively associated with alcohol (r = 0.216, p<0.0001) and drug use (r = 0.225, p<0.0001) and negatively associated with increasing age (r = -0.088, p- = 0.02) and female gender (r = -0.232, p<0.0001). Factor analyses repeated in the second random pre-campaign subsample and post-campaign sample confirmed these results. Conclusion: A vignette-based scale may be an effective measure of key attitudes related to concurrency and potentially a useful tool to evaluate interventions addressing this network pattern. [ABSTRACT FROM AUTHOR]