 |
Center for Nonlinear Studies |
The following lecture is part of the Invited Statistical Sciences Seminar Series (CCS-6).
Assessing educational initiatives that look to improve student success and their effect on specific populations of students is a growing goal of many college campuses. In this talk, we extend interaction trees to be used with data available from observational studies. Simulation studies will be presented where the random forest of interaction trees methods developed are compared to alternative routines on its ability to provide individual estimates of, determine the variables most associated with, and create subgroups of students based on performance improvement in a particular modality, intervention or initiative. An application to an optional recitation course for a large lecture introductory statistics course will be presented. The proposed method and framework allow for individualized advising of students based on their predicted treatment effect in a particular modality.
Host: Kary Myers, kary@lanl.gov