Research Experiences For Undergraduates (REU)

REU 2008 Projects

A Macroscale Model for the Growth of Glioblastoma Multiforme Incorporating Tumor Cell Heterogeneity

Sponsor: IBM
Advisor: Prof. Suzanne Weekes

David B. Brown    Stephanie Browne-Schlack    Thomas Seaquist    Erin Vinnedge
David B. Brown    Stephanie Browne-Schlack    Thomas Seaquist    Erin Vinnedge

Glioblastoma multiforme (GBM) is a highly lethal brain cancer--patients typically survive 12-18 months after diagnosis with treatment, and only six months without. Our continuum model describes the dynamics of GBM via a system of nine partial differential equations involving tumor cell, nutrient, toxin, and chemoattractor densities. We allow for genetic instability of the tumor cells by considering five different tumor cell clone types which may mutate upon proliferation. The mutations follow a linear progression pathway from clone A (the least aggressive type) to clone E (the most aggressive type). Since our model is macroscopic in nature, a key objective is to obtain results similar to those produced by previous agent-based models set on a microscale. Using numerical methods, we approximate the solution to our model in MATLAB and conduct in-silico experiments.

Disposition Effect and Momentum: Modeling Irrational Behavior to Explain Apparent Momentum in Stock Prices

Sponsor: State Street Global Advisors (SSgA)
Advisor: Prof. Marcel Blais

David P. Brown    Pulak Goswami    Xiao Han    Evan Herring    Jian Wang
David P. Brown    Pulak Goswami    Xiao Han    Evan Herring    Jian Wang

Grinblatt and Han (2005) model momentum using the frameworks of Prospect Theory and Mental Accounting (PT/MA). These frameworks explain the tendency of some investors to hold onto losing stocks too long and sell winning stocks prematurely, resulting in price distortions in the market which appear as momentum as market forces act to correct these discrepancies. Using stock market data provided by State Street Global Advisors (SSgA), we compare Grinblatt and Han's model with some traditional, empirically motivated measures of momentum. We utilize the information coefficient (IC), the correlation between a model's prediction and actual outcomes, as a metric of testing success. Although results show that the PT/MA model has some predictive power, the model does about the same, if not worse than traditional momentum measures at forecasting stock movements, leading us to conclude that PT/MA based momentum should not replace traditional measures of momentum when forecasting stock prices.

The Effect of Missingness on the Analysis of Alzheimer's Disease Clinical Trials of Disease Modifying Agents

Sponsor: Pfizer
Advisor: Prof. Jayson Wilbur

Sara Adler    Jennifer Boyko    Laura Hou
Sara Adler    Jennifer Boyko    Laura Hou

Patients often withdraw from clinical trials for various reasons, including adverse health effects, lack of drug efficacy, and death. The resulting missing data poses a problem for analysis of the data. The goal of this project was to understand the effect of various types of missing data on the analysis of a clinical trial for a degenerative disease such as Alzheimer's Disease. Data were simulated based on the expected longitudinal progression of a clinical trial for an Alzheimer's Disease Disease Modifying (ADDM) drug, with varying degrees of variation and missingness. Missingness was simulated as Missing Completely at Random (MCAR), Missing at Random (MAR) and Missing Not at Random (MNAR). Analysis of these simulated data showed that complete cases analysis and multiple imputation (MI) are more appropriate for this type of clinical study than the standard Last Observation Carried Forward (LOCF) approach, yielding higher power and lower bias. We present recommendations for use of specific methods for various combinations of the standard deviation, degree of missingness, type of missingness, and other parameters.

Maintained by
Last modified: Jun 20, 2010, 23:10 EDT
[WPI] [Math] [Home]