Good Things Come to Those Who Weight? Testing Weighting Necessity Metrics for Complex Survey Data

Citation

Brehm, Christopher K. (2023). Good Things Come to Those Who Weight? Testing Weighting Necessity Metrics for Complex Survey Data.

Abstract

I studied the performance of four statistical tests designed to assess the necessity of weighting in the analysis of survey data. First, I investigated the finite sample properties of diagnostic tests using Monte Carlo simulations, including a case in which weights are ignorable and a case where weights are designed to be necessary. I tested the performance of four necessity metrics across both scenarios. Second, I took a sample of studies using data and weights from Add Health. I replicated models proposed in each study and applied weighting necessity diagnostic tests. Results from the simulation study found that the DuMouchel-Duncan test exhibited the strongest performance most consistently. Demonstrating the use of weight necessity tests on replicated models revealed that Difference in Coefficient Tests returned significant test statistics most often. These results offer insight into the selection of appropriate weight necessity tests, contributing to more accurate analysis in complex survey settings.

URL

https://www.proquest.com/docview/2912882041?pq-origsite=gscholar&fromopenview=true&sourcetype=Dissertations%20&%20Theses#

Keyword(s)

Adjustment error

Reference Type

Thesis/Dissertation

Book Title

Sociology

Author(s)

Brehm, Christopher K.

Series Author(s)

Bollen, Kenneth A.

Year Published

2023

Volume Number

Masters

Publisher

The University of North Carolina at Chapel Hill

City of Publication

United States -- North Carolina

ISSN/ISBN

9798381384161

DOI