A Re-Evaluation of the Statistical Learning Approach to Optimal Sample Allocation

• Activity Number: 606
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Ismael Flores Cervantes*
• Companies: Westat
• Keywords: Optimal Sample allocation ; statistical learning ; Neyman allocation ; imperfect design data

Abstract

Statistical learning is both a theory and a group of algorithms for machine learning. These algorithms detect and learn patterns in data that, in turn, can be used to predict outcomes. These methods have become very popular in recent years and have been successfully applied in many fields. One criticism is that these methods are mainly used as black boxes without a good understating of how they work. Clark (2013) developed a novel approach to optimal sample allocation in stratified design based on a statistical learning approach (SL). However, the SL sample allocation is not without problems when compared to simpler allocations. In this paper we expand the research on the SL approach and re-evaluate its performance. We describe the mechanism behind the SL sample allocation for situations not previously considered.