Button to scroll to the top of the page.

Summer 2019 Colloquium: Graduate Portfolio in Applied Statistical Modeling






Ahmed Hassan

July 29


GDC 7.402

"Vectorized Cerebrovascular Network Curation"
Chun Liu Aug. 20 9:30-10am GDC 7.402 "A multi-level analysis of the effects of Independent Living Programs on educational attainment, employment, and housing outcomes of youth aging out of foster care"
Prageet Kang Aug. 20 11-11:30am GDC 7.402 "Zinc, Copper, and Iron in Oral Submucous Fibrosis: A Meta-Analysis"

Ahmed Hassan

Title: "Vectorized Cerebrovascular Network Curation"

Abstract: Stroke is a shockingly common event that can lead to critical, chronic disability. In order to characterize the extent and severity of ischemic events, as well as recovery post-ischemia, cerebrovascular imaging is essential. Resulting image data imposes a large computational burden and is difficult to analyze quantitatively, meaning that vectorization is required to transform the multiphoton images into structures that can be readily quantified and manipulated. However, vectorization approaches are susceptible to errors and all outputs must be curated. Manual curation is infeasible due to the enormous number of strands that occupy a mere cubic millimeter volume of cortex. Thus, an automated classifier was developed. More than 2,000 strands were manually classified for ground truth, and strand features describing the geometric, microscopic, and vascular characteristics of each strand were extracted. Exploratory data visualization was performed to gain insight on which features may function as viable classification predictors, and to guide troubleshooting efforts. Subsequently, interrelated and non-significant features were removed using multicollinearity analysis and backwards feature elimination respectively. Two learning algorithms were trained using the remaining feature set: binary logistic regression and Adaboost. Model evaluation via receiver operating characteristics area-under-the-curve showed decent performance (AUC > 0.85) without overfitting, as evidenced by similar performance on test and train fractions. Finally, data curation using a single decision threshold (fully automated curation) or multiple decision thresholds (semi-automated curation requiring manual intervention) was evaluated. For both approaches, Adaboost was shown to outperform binary logistic regression. For fully automated curation, logistic regression results in a ~0.78 classification accuracy relative to Adaboost’s modestly larger ~0.805. Using multiple decision thresholds to achieve a 0.90 classification accuracy, the logistic regression classifier required ~30% of strands to be manually curated compared to Adaboost, which required ~20% be subject to manual intervention. While these results are promising, the inclusion of a more robust initial feature set should improve model performance, and future efforts should examine how well trained classifiers perform across healthy and post-ischemic specimens.


Chun Lui

Title:  "A multi-level analysis of the effects of Independent Living Programs on educational attainment, employment, and housing outcomes of youth aging out of foster care."

Abstract: Background: The transition from adolescence to adulthood is a crucial period in a young person’s life. It is even more challenging for foster care youth. Every year there were about 20,000 youths aging out of foster care system. They normally face multiple disadvantages in terms of educational attainment, employment, housing, financial stability, and life skills compared with children in the general population. About two-thirds of eligible foster youth receive Independent Living Programs (ILPs), which are designed to support youth to assure a successful transition to adulthood. The objective of this paper is to examine whether ILPs are effectively promoting better outcomes (e.g. educational attainment, employment, housing) for youth aging out of foster care. Methods: Using data from the National Youth in Transition Database (NYTD), this study used Hierarchical Generalized Linear Models (HGLM) to investigate how different services from ILPs impact the educational attainment, employment, and housing outcomes of youth aging out of foster care across all 50 states. The study sample includes foster youth from FY 2014 cohort (N = 5633) on Wave 3 at age 21. The dependent variables are educational attainment, employment, and housing outcomes. The independent variables are services provided by ILPs in the following areas: special education, independent living needs assessment, academic support, post-secondary educational support, career preparation, employment programs or vocational training, budget and financial management, housing education and home management training, health education and risk prevention, family support and healthy marriage education, mentoring, supervised independent living, room and board financial assistance, education financial assistance, other financial assistance. The covariates include youth gender, race, delinquency, foster care status. Results: Controlling for all the explanatory variables and the random effect, youth who received post-secondary educational support (OR = 1.718, p < .001), budget and financial management (OR = 1.242, p = .038), and education financial assistance (OR = 1.440, p = .003) were more likely to achieve higher educational attainment. Youth who received post-secondary educational support (OR = 1.427, p < .001) and supervised independent living (OR = 1.288, p = .024) were more likely to get employed. Conclusions and Implications: The results indicate that some certain types of ILPs services are associated with positive outcomes in terms of education and employment. Post-secondary educational support service is found to be the most effective type of service for improving both the educational attainment and employment outcomes. The findings suggest the importance of providing ILPs to youth aging out of foster care. In addition, variation in service delivery and implementation fidelity across states must be taken into consideration.


Prageet Kang

Title: "Zinc, Copper, and Iron in Oral Submucous Fibrosis: A Meta-Analysis"

Abstract: Oral submucous fibrosis (OSF) is a potentially malignant disorder which causes fibrosis and inflammation of the oral mucosa. Studies have reported altered levels of trace elements in oral submucous fibrosis subjects, but findings have been inconsistent. The objective of this research is to perform a meta-analysis to summarize studies that report zinc (Zn), copper (Cu), and iron (Fe) in patients, with and without OSF. A literature search of Embase, PubMed, Cochrane Library, and Web of Science electronic databases was conducted for studies up to January 2017. A total of 34 reports met the inclusion criteria. The standardized mean difference was utilized as the effect size. The robust variance estimation method was chosen to handle dependency of multiple related outcomes in meta-analysis. There was a significant increase in the levels of Cu (effect size=1.17, p value < 0.05, 95% confidence interval (CI): 0.164–2.171) and a significant decrease in levels of Zn (effect size= −1.95, p value < 0.05, 95% CI: −3.524 to −0.367) and Fe (effect size= −2.77, p value < 0.01, 95% CI: −4.126 to −1.406) in OSF patients. The estimation of Zn, Cu, and Fe levels may serve as additional biomarkers in the diagnosis and prognosis of OSF along with the clinical features.