A wide range of missing imputation approaches in longitudinal data: a simulation study and real data analysis.

Area of Science:

• Statistics
• Biostatistics
• Data Science

Background:

• Missing data is a significant challenge in longitudinal data analysis.
• Various single-imputation (SI) and multiple-imputation (MI) methods exist to address missing data.
• The efficacy of non-parametric methods for imputation in longitudinal studies requires further investigation.

Purpose of the Study:

• To investigate the performance of the longitudinal regression tree algorithm as a non-parametric method for imputing missing data in longitudinal studies.
• To compare the effectiveness of 27 different imputation approaches (SI and MI) using both simulated and real-world data.
• To assess imputation performance using metrics like Mean Squared Error (MSE), Root-Mean-Squared Error (RMSE), and Median Absolute Deviation (MAD).

Main Methods:

• Compared various imputation methods including cross, trajectory mean, interpolation, copy-mean, and MI.
• Utilized simulated data scenarios and real data from the Tehran Cardiometabolic Genetic Study (TCGS) with 3,645 participants.
• Modeled systolic and diastolic blood pressure (SBP/DBP) using predictor variables like age, gender, and BMI, employing both parametric and non-parametric longitudinal models.

Main Results:

• The longitudinal regression tree algorithm demonstrated superior performance over the linear mixed-effects model (LMM) based on MSE, RMSE, and MAD criteria for both TCGS and simulated data under a Missing At Random (MAR) mechanism.
• The performance of the 27 imputation approaches was largely similar when fitting the non-parametric model.
• The single-imputation trajectory mean (SI traj-mean) method showed improved performance compared to other imputation approaches.

Conclusions:

• Both SI and MI imputation approaches performed better with the longitudinal regression tree algorithm than with parametric longitudinal models.
• Researchers are recommended to use the trajectory mean (traj-mean) method for imputing missing values in longitudinal data.
• The optimal imputation method selection depends on the specific models and data structure of interest.