American Statistical Association
An outgrowth of both targeted cancer therapy and improved measures of tumor biology is the clinical demand and justification to investigate subgroup prognosis or interaction effects with treatment. To facilitate use of the results for future clinical trial design, the statistical methods we will investigate allow investigator control over the class of decision rules and the fraction of patients identified by the rules.
A variety of data partitioning methods have proven useful for constructing interpretable descriptions for groups of patients based on decision rules. We explore a new method called Extreme Regression for finding patient subsets corresponding to either very good or very poor treatment outcomes. The regression method specifies a model class composed of extrema (maximum and minimum) functions of the predictor variables. This class of models allows for simple function inversion and results in level sets of the regression function which can be expressed as interpretable decisions based on individual predictors. Some results based on recently conducted clinical trials for Non-Hodgkin's lymphoma and multiple myeloma are presented.
|Date:||Wednesday, November 7, 2007|
|Time:||4:00 P.M. - 5:00 P.M.|
Memorial Sloan-Kettering Cancer Center
Department of Epidemiology and Biostatistics
307 East 63rd Street
(between First and Second Avenues)
3rd Floor Conference Room
New York, New York
Note: To gain access to the building, please follow the directions by the telephone in the foyer.