Generalized Boosted Models to Measure Racial Effects at Different Quantiles in Observational Studies

Area of Science:

• Epidemiology
• Biostatistics
• Longitudinal Data Analysis

Background:

• Observational studies with longitudinal data present challenges in estimating treatment effects, particularly at different quantiles.
• The National Heart, Lung, and Blood Institute (NHLBI) Growth and Health Study (NGHS) provides a relevant dataset for examining racial effects on cardiovascular risk factors over time.

Purpose of the Study:

• To develop and evaluate a novel estimation method for quantile treatment effects in observational longitudinal studies.
• To assess racial disparities in cardiovascular risk factors using the NGHS data by estimating effects at various quantiles.

Main Methods:

• Employed a nonparametric generalized boosted models (GBM) approach to estimate unknown propensity score models.
• Developed a GBM-based quantile weighting estimation method by integrating quantile regression and inverse probability weighting.
• Applied the proposed method to NGHS data to quantify racial effects across different quantile levels.

Main Results:

• The study found that racial effects on cardiovascular risk factors are not uniform and vary significantly across different quantile levels.
• Results indicate that racial effects may be non-zero, highlighting potential disparities that are missed by traditional average treatment effect analyses.

Conclusions:

• The proposed GBM-based quantile weighting method is effective for estimating quantile treatment effects in longitudinal observational studies.
• The findings underscore the importance of considering quantile-specific effects to fully understand complex relationships, such as race and cardiovascular health.