Dissertation Defenses

Upcoming Dissertation Defenses

Name:  Tarek A. Gad

TitleElasticity Theory of Learning Growth in the 21st Century

Committee Chair:  Dr. George Karabatsos

Date:  Thursday, October 11, 2018

Time:  10:00am – 12:00 pm

Location:  Room 3015, ETMSW Building, 1040 W. Harrison St.


The challenge that faces the student for the 21st century is enormous, and requires the collective effort of all the stakeholders to improve academic outcomes for students. My dissertation provides a framework that is based on the ecological framework to understand the dynamic interrelations among various personal and environmental factors. The theory that I have developed in this dissertation has the potential to include all the variables (+6000 variables) that the NELS/ELS: 2002 contains. The NELS:2002 data includes surveys from students, teachers, parents, and administrators in a series of data collection.

The Meinshausen-Bühlmann, MB, algorithm (high-dimensional graph model) selects the variables that can predict a target variable of choice through a lasso regression process. The MB algorithm produces a graph that demonstrates the conditional dependence and independence across all the variables under study. In order to connect funding to learning, the elasticity theory analysis will provide guidelines in the process of selecting the elements that have the highest return on investment.

The results of this study revealed that a mother occupation, a student gender, a student self-confidence, a grandparents’ education, an English native language, a reading proficiency, prior knowledge, a parent’s expectation, and a family SES directly impact student academic achievement. However, family composition, parents’ education, family income, student ethnicity, teacher education, and principal leadership indirectly influences the student performance.

The framework in this dissertation provides a broad scale of data analysis and different approaches to interpret statistics based on the variable’s elasticity. The theory in this dissertation provides a new approach to the analysis of complex data (complex data is a data that is hard to analyze with traditional approaches) such as the NELS:2002 (+6000 variables, and 16179 entries). This new approach will change the traditional data analytics landscape across all industries especially in education by including the elasticity theory as an additional factor to interpret the statistical results.


Name:  Nicole K. Ozturk

TitleA Bayesian Robust IRT Outlier Detection Model

Committee Chair:  Dr. George Karabatsos

Date:  Thursday, October 18, 2018

Time:  3:30pm – 5:30 pm

Location:  Room 3015, ETMSW Building, 1040 W. Harrison St.


In the context of psychometric practice, the parameter estimates of a standard item-response theory (IRT) model may become biased when item-response data, of persons’ individual responses to test items, contain outliers relative to the model. Further, the manual removal of outliers can be a time-consuming and difficult task. Besides, removing outliers leads to data information loss in parameter estimation. To address these concerns, a Bayesian IRT model that includes person and latent item-response outlier parameters, in addition to person ability and item parameters, is proposed and illustrated, and defined by item characteristic curves (ICCs) that are each specified by a robust, Student’s t-distribution function. The outlier parameters and the robust ICCs enable the model to automatically identify item-response outliers, and to make estimates of the person ability and item parameters more robust in the presence of outliers. Hence, under this IRT model, it is unnecessary to remove outliers from the data analysis.

The Bayesian IRT model is illustrated through the analysis of two real-world, and two simulated datasets involving dichotomous- and polytomous-response items. Additionally, the model is applied to a simulated skewed dichotomously scored assessment to more closely understand how the model performs under realistic testing conditions.