Projects

Synergistic Activities at PRIISM

High-Dimensional Data Compression/Feature Extraction applied to Kinematic Data

Collaboration Between Dr. Preeti Raghavan (Motor Recovery Lab at Rusk Institute), and Dr. Ying Lu (member of PRIISM)

Using rich information of kinematic and EMG data collected at the Motor Recovery Lab, we are interested in the movement patterns and how they change when the physiology is modified due to training, injury, disease and disability. Statistical models for measuring change in movement patterns after interventions is desired. Dimensionality reduction, functional data analysis and dynamic factor analysis models are being explored and developed.

New Psychometric Models for the Assessment of Cognitive Skills Using Collaborative Problem Solving Tasks

Collaboration Between Dr. Alina A. von Davier (Educational Testing Service), and Dr. Peter Halpin (member of PRIISM)

Statistical modeling of incidental data from learning technologies presents some interesting opportunities. In context of collaboration, activity logs provide a time-stamped record of how students communicate and interact with one another while solving a complex problem. These time-stamped events can be modeled as a point process.  In comparison with traditional psychometric models, which only condition on the ability of the single individual being assessed, point processes also condition the previous activities of all the other group members. Intuitively, this means that we reckon on the (demonstrated) ability of the entire group when estimating the (latent) ability of each specific individual. Point processes also serve to incorporate the timing of activities, whereas traditional models for assessment have only considered whether activities are "correct" or not. The development of appropriate psychometric models, algorithms for their estimation, and software for the assessment and data capture of collaborative problem solving tasks is the focal topic of this research.

Sensitivity Analysis for Causal Inference

Jennifer Hill, Marc Scott, Nicole Carnegie & Joel Middleton (all of PRIISM).

Education researchers often find that available data are not well suited for answering the causal questions they are most interested in answering. This can even be the case when they have collected the data themselves. Randomized experiments are not always feasible due to logistical or ethical concerns. Even when data from randomized experiments are available, such experiments are often “broken” due to missing follow-up data or non-compliance with treatment assignment. In these situations researchers may be forced to make strong, often untestable assumptions in order to make causal inferences. Although researchers cannot always empirically determine whether or not these assumptions hold, they should be able to use their data, along with their substantive knowledge of the area, to gauge how far estimates might be from the truth by performing sensitivity analysis. Funded by The Institute for Education Sciences, US Dept. of Education.

The Cancer, Insulin Resistance and Lifestyle (CIRCLE) Study in the Framingham Heart Study Population

Collaboration Between Niyati Parekh (NFS&PH), and Marc Scott (PRIISM). Research Assistance: Maya Vadiveloo

The overall purpose of the study titled is to investigate the interrelationships of physiologic, dietary and genetic factors associated with disturbances in the insulin-glucose axis in relation to combined and site-specific incidence of obesity-related cancers, by performing secondary data analyses in a large existing sample of American adults from the National Heart, Lung, and Blood Institute’s (NHLBI) Framingham Heart Study (FHS;1948-2008) consisting of ~14,000 adults 20 years or older. There is sufficient evidence in the literature to support the obesity-cancer link. Although potential biological mechanisms of obesity-related metabolic abnormalities on cancer risk have been hypothesized and confirmed by laboratory studies these relationships have not yet been fully characterized in humans, and remain unclear; this is a primary area of enquiry in the CIRCLE study.