Triangulation, Modeling and Adjustment : Correcting Online Surveys for Coverage Biases Among Non-Internet Households Using Machine Learning Methods Trained on Multiple Data Sources Spanning Several Domains / David Dutwin, Trent D. Buskirk.

Format:

Book

Author/Creator:

Dutwin, David, author.

Buskirk, Trent D., author.

Language:

English

Subjects (All):

Machine learning.

Internet surveys.

Physical Description:

1 online resource

Place of Publication:

London : SAGE Publications Ltd, 2025.

Summary:

Over the past 15 years survey research has moved strongly toward online interviewing, much of which only samples from persons with internet access. Unfortunately, while the non-internet population is small, many behavioral and attitudinal differences between those that use and do not use the internet are acute, leading to potentially significant bias in survey estimates. As reported in the referenced article, "A Deeper Dive on the Digital Divide: Reducing Coverage Bias in Internet Surveys," we set out to develop an adjustment model that could be utilized to minimize bias due to noncoverage of the non-internet population. There were three challenges to our work that this case study details. First, where could we find high quality microdata with a consistent metric for internet use along with other metrics with which to correlate against internet use? Second, assuming potentially hundreds of indicators across many datasets, how could we reduce such data down to those variables most predictive of internet use, and how would we deal with any collinearity and duplication across variables? Finally, how could we develop a single unified model given variables spanning multiple data sources? This case study explores these issues in detail and provides our answers to these challenges, hoping to help others facing similar challenges.

Notes:

Description based on publisher supplied metadata and other sources.

ISBN:

1-03-621559-8

9781036215590

OCLC:

1523169708