Debiased machine learning for logistic partially linear mediation models with high-dimensional confounders

Key Points

• Mediation effects are effectively estimated using debiased machine learning techniques, indicating improved model accuracy.
• The approach applies in scenarios with high-dimensional confounders, revealing complex relationships between variables.
• Assessment uses novel logistic partially linear mediation models, showcasing flexibility in data analysis.
• Findings support the need for robust statistical methods in high-dimensional research settings, addressing limitations of traditional models.