Incorporating the sampling design in weighting adjustments for panel attrition

Area of Science:

• Survey Methodology
• Statistical Analysis

Background:

• Panel attrition, the non-response of participants over time, can introduce significant bias into survey estimates.
• Traditional weighting adjustment methods may not fully account for complex survey designs.

Purpose of the Study:

• To review and suggest methods for weighting adjustments that incorporate survey design variables to address panel attrition.
• To demonstrate the effectiveness of these methods in reducing bias and maintaining weight stability.

Main Methods:

• Review of existing weighting adjustment techniques for panel attrition.
• Incorporation of design variables (strata, clusters, baseline weights) into attrition analysis.
• Utilizing multilevel models and decision tree algorithms (e.g., chi-square automatic interaction detection).
• Simulation studies to evaluate the performance of proposed methods.

Main Results:

• Weighting approaches incorporating design factors effectively reduce attrition bias in survey estimates.
• These methods help maintain stable resulting weights.
• Simulation results confirm the efficacy of the proposed techniques.

Conclusions:

• Integrating survey design information into attrition weighting adjustments is crucial for accurate survey estimates.
• The proposed methods offer practical solutions for analysts dealing with panel attrition in complex surveys.
• A case study illustrates the application of these techniques in a post-disaster community survey.