| What? | micro Big Data Analytics workshop. The focus of this micro-workshop is to study the foundations for developing a new Compressive Big Data Analytics (CBDA) foundation enabling representation, modeling, analysis and interrogation of large, incongruent multi-source, incomplete and messy data. The highlights of the workshop include talks from Dr. Saeid Amiri and Dr. S. Ejaz Ahmed. |
| Presenter | Dr. S Ejaz Ahmed (Brock University) |
| Title | BIG DATA: Variable selection, post estimation and prediction. |
| Where | TBD |
| Date | Friday, April 24, 2015 |
| When | 9 AM (TBD) |
| Abstract | In high-dimensional data analysis it is commonly assumed that the predictive model is sparse. However, assumed model may have a number of weak signals together strong and sparse signals. Consequently, aggressive variable selection techniques may eliminate weak signals from the model to select a parsimonious submodel. However, the prediction accuracy based on such submodel may not be desirable in such casses. For this reason, we propose high-dimensional shrinkage estimation strategies to improve the prediction performance of the submodel. The relative performance of the proposed prediction strategies are evaluated both analytically and by simulation studies. Two real data examples are presented to illustrate the proposed methodologies. |
| See also | Shrinkage, pretest, and penalty estimators in generalized linear models (10.1016/j.stamet.2014.11.003). |
| To get engaged or meet with the speaker contact | Ivo Dinov. |