## Industrial Projects

# 2005-06 Projects

### Mathematical Modeling of Torque for Screw Insertion Process

**Sponsor: BOSE Corporation
Advisor: Suzanne L. Weekes
Students: Angela A. Leo, Sanjayan Manivannan, John R. Potter**

A self-tapping screw is a high-strength one-piece fastener that is driven into preformed holes. We analyze and improve a mathematical model of the self-tapping screw insertion process so that it can be used in manufacturing processes at the BOSE Corporation. We build a Graphical User Interface in MATHLAB which allows users to enter fundamental data and produces the corresponding torque curve. The accuracy and robustness of the model is tested by comparing predictions to empirical data collected at BOSE.

### Expert System Design for Long-Term Care Underwriting

**Sponsor: John Hancock
Advisor: Jon P. Abraham
Students: Elizabeth A. Arsenault, Nicholas E. Rackliff**

The project presents a mathematical model created to determine a risk score for applicants for Long-Term Care insurance. We developed software to assist in the implementation and testing of the model. We analyze the suitability of our model, including the sensitivity of scores to the model's parameters. We formulate methods which could help train the model, including numerical regression to solve for model parameters and pair-wise comparisons between applicants to verify consistency of parameters and risk factors.

### Lehman Brothers: Risk Reporting and Technology

**Sponsor: Lehman Brothers
Advisors: Arthur Gerstenfeld, Michael J. Ciaraldi, Jon P. Abraham
Students: William Braden Hays, Amy Louise Jackson, Jason T. Tondreau, Igor Ushakov**

The main goal of this project was to improve risk reporting and technology at Lehman Brothers for both country and credit risk. The project was divided into two separate parts that were completed in parallel. One part was completed in New York which focused on the improvement of the country risk reporting process. The other was executed in London to consolidate foreign-exchange spot credit limits.

### Robustifying Logistic Regression: An Application to Obesity

**Sponsor: National Center for Health Statistics/CDC
Advisor: Balgobin Nandram
Student: William Marjerison**

We predict finite population mean BMI nationally for children and adolescents using NHANES III survey data. There are many nonrespondents and no distributional assumption is made on BMI. As link functions for response indicators, we compare the logistic distribution assumption is made on BMI. As link functions for response indicators, we compare the logistic distribution and student's mixtures. Nonrespondents are assigned cells based on propensity scores to impute BMI, and uncertainty about this process is included. Predictive inference is done using least-squares, and we compare results with a recent method.

Maintained by webmaster@wpi.eduLast modified: Jun 20, 2010, 09:03 EDT