SeminarsPrediction and Diagnosis of Disease Outcome with Microarray Data and a Classification Method Based on a New Information Measure
140
reads
The advancement of technology has made high dimensional data available in an ever-growing rate. These data can be used to answer significant questions. Thus finding a small subset of most predictive variables and use them effectively in prediction is the new challenge in many disciplines. In this talk, we employ a new information measure introduced in Lo and Zheng (2002) and develop an algorithm to predict the disease outcome based on microarray data. The new information measure can assess the total information content of a set of variables including interactions among variables and exhibit appropriate sensitivity when dropping a relevant or an irrelevant variable from the set of variables regardless the form of relationship between the response and explanatory variables. We also report encouraging results from a challenging microarray data set for classifying breast cancer.
reads
Inchi Hu
2009-01-07
15:30:00 - 17:00:00
308 , Mathematics Research Center Building (ori. New Math. Bldg.)
The advancement of technology has made high dimensional data available in an ever-growing rate. These data can be used to answer significant questions. Thus finding a small subset of most predictive variables and use them effectively in prediction is the new challenge in many disciplines. In this talk, we employ a new information measure introduced in Lo and Zheng (2002) and develop an algorithm to predict the disease outcome based on microarray data. The new information measure can assess the total information content of a set of variables including interactions among variables and exhibit appropriate sensitivity when dropping a relevant or an irrelevant variable from the set of variables regardless the form of relationship between the response and explanatory variables. We also report encouraging results from a challenging microarray data set for classifying breast cancer.