|
Activity Number:
|
285
- New Advances in Sample Design and Adjusting for Survey Nonresponse
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Survey Research Methods Section
|
|
Abstract #318977
|
|
|
Title:
|
Evaluation and Utility of Address-Level Predictive Models for Address-Based Sampling (ABS) Sample Designs
|
|
Author(s):
|
Joseph McMichael* and Stephanie Zimmer
|
|
Companies:
|
RTI International and RTI International
|
|
Keywords:
|
ABS;
stratification;
sample design;
auxiliary data
|
|
Abstract:
|
The use of auxiliary data in sample design is becoming more common on surveys using Address Based Sampling (ABS) frames because of the ability to link variables to frame elements through an address or location. These auxiliary data can be used for stratification, decisions about data collection protocols, and weight adjustments. Auxiliary data from both public and private sources can be flawed or imprecise and is available at different levels. Public data is commonly available by census geography while privately sourced data is at the address or person level. For sample design, address-level data about demographics or rare characteristics is desirable. Rather than using flawed data directly, we improve its utility by creating address-level models for the characteristics of interest. This paper discusses the development of address-level models that predict demographic characteristics for survey design. Models were trained on three ABS surveys using predictors from public and private data. We discuss model development including variable selection and data transformation. We also evaluate portability: how well a model independently trained on one survey applies to a different survey.
|
Authors who are presenting talks have a * after their name.
