Department of Quantitative Health Sciences
Lerner Research Institute, Cleveland Clinic
“Variable Screening, Selection and Prediction in High Dimensions”
We consider an alternative method for variable screening, selection and prediction in linear regression problems where the number of predictors are much higher than the number of observations. The method involves minimizing a penalized Euclidean distance with a newly defined penalty term. This particular formulation exhibits a grouping effect, which is useful for screening out predictors in higher or ultra-high dimensional problems. Practical performances of variable selection and prediction are evaluated through simulation studies and the analysis of a dataset of mass spectrometry scans from melanoma patients, where excellent predictive performance is obtained.
Friday, April 28, 2017, 11:30 AM, BLOC 113