A hierarchical model for binary data with dependence between the design and outcome success probabilities.

Area of Science:

• Biostatistics
• Health Disparities Research
• Epidemiology

Background:

• Statistical analysis requires conditioning on the study design process to prevent biased estimates.
• Ignoring design-outcome associations can lead to erroneous inferences, particularly in health research.
• Understanding factors influencing health literacy is crucial for targeted interventions.

Purpose of the Study:

• To propose hierarchical models investigating the dependence between study design and outcomes.
• To compare fully parametric and semi-parametric model formulations.
• To gain insight into breast cancer literacy mechanisms among underserved females.

Main Methods:

• Development of a class of hierarchical models.
• Formulation of both fully parametric and semi-parametric models.
• Application of the Expectation-Maximization (EM) algorithm for maximum likelihood estimation.

Main Results:

• The semi-parametric model demonstrated superior performance compared to the fully parametric model.
• Identified key features where the semi-parametric approach excelled.
• Provided insights into the generation of breast cancer literacy outcomes.

Conclusions:

• Accounting for the design process in statistical analysis is essential for valid inferences.
• Semi-parametric hierarchical models offer a robust approach for studying design-outcome dependencies.
• The findings contribute to understanding and addressing breast cancer literacy disparities in vulnerable populations.