|
Activity Number:
|
99
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #303327 |
|
Title:
|
An Introduction to Pre-sampling Inference
|
|
Author(s):
|
Stephen Woodruff*+
|
|
Companies:
|
Specified Designs
|
|
Address:
|
800 West View Terrace, Alexandria, VA, 22301-2750,
|
|
Keywords:
|
Design Based Estimation ; Model based Estimation ; Pre-sampling ; Linear Models
|
|
Abstract:
|
Efficient probability sample design is inhibited by operational constraints and design based selection-probability estimators, although unbiased, are seldom variance-efficient. Model based estimators are both minimum variance and unbiased but rely on models gleaned from potentially misleading sample data. This paper extends model based inference to models imposed on sample data by designed randomization. This combines the advantages of randomization with the inferential power of models under statistical warranty. This is done by random construction of the sample units called pre-sampling. In the multivariate application described here, the model based BLUE has a variance that is orders-of-magnitude smaller than that of design based competitors. Pre-sampling methodologies appear to be robust against rough estimation of the model parameters and are evaluated under repeated sampling.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
