Colloquium
- Home
- About
- Academics
- Courses
- Undergraduate
- Graduate
- Alumni
- Career Guidance
- Giving
- Research
- Resources
- Stat Consulting Center
- Directory
- Contact Us
Statistics Colloquium/Research Proposal
November 22, 11 a.m.
Alavi Commons Room, 6625 Everett Tower
Ida Alcantara
PhD Candidate, Department of Statistics
Abstract:
The coefficient of determination and other statistics based on SSE
cannot be used for variable selection because SSE can only decrease when adding predictors.
We explore alternative statistics based on PRESS, or 'predictive residual sum of squares'.
Unlike SSE, the PRESS statistic can actually increase with the presence of irrelevant
predictors, and hence could be a diagnostic that a smaller model is necessary.
We will also examine properties of PRESS that may be useful in inference. For example, we propose a PRESS-based unbiased estimator of error term variance as an alternative to MSE. We will explore situations under which the PRESS-based estimator might be preferable.
Bio:
Ida Alcantara is a PhD student in the Department of Statistics.
---
All statistics graduate students are expected to attend.