A Bayesian semiparametric approach for inference on the population partly conditional mean from longitudinal data...

Area of Science:

• Cognitive Neuroscience
• Biostatistics
• Longitudinal Data Analysis

Background:

• Longitudinal memory studies often yield biased results due to non-representative samples from selective enrollment and attrition.
• Practice effects, stemming from test familiarity, further distort memory trajectory findings and limit generalizability.
• Accurate population-level inference for longitudinal outcomes is challenging.

Purpose of the Study:

• To develop and validate a novel statistical approach for estimating population means of longitudinal outcomes, accounting for mortality.
• To address biases inherent in longitudinal cohort studies, particularly in memory research.
• To generalize findings from specific cohorts to the broader target population.

Main Methods:

• A Bayesian semiparametric predictive estimator was developed for population inference using longitudinal auxiliary data.
• The method estimates finite population means conditional on survival at a specific time point.
• Sensitivity analyses and simulation studies were conducted to assess robustness and compare with existing methods.

Main Results:

• The proposed method demonstrated improved accuracy in estimating population-level memory trajectories compared to traditional approaches.
• Bias due to selective attrition and practice effects was effectively mitigated in simulation studies.
• The approach proved robust under various untestable assumptions.

Conclusions:

• The developed Bayesian approach offers a robust solution for population inference in longitudinal studies with selective attrition and mortality.
• Accurate estimation of lifespan trajectories in cognitive functions like episodic memory is achievable, enhancing generalizability.
• This method provides a valuable tool for researchers aiming for unbiased population-level insights from cohort data.