The performance of mixed truncated spline-local linear nonparametric regression model for longitudinal data

Area of Science:

• Statistics
• Econometrics
• Biostatistics

Background:

• Nonparametric regression models handle complex data patterns.
• Mixed estimator nonparametric regression (MENR) combines multiple estimators for multivariable analysis.
• Longitudinal data presents unique challenges due to repeated measurements over time.

Purpose of the Study:

• To develop a novel mixed truncated spline-local linear nonparametric regression (MTSLLNR) model for longitudinal data.
• To address situations where predictor variables exhibit differing data patterns.
• To evaluate the performance and consistency of the proposed MTSLLNR model.

Main Methods:

• The study proposes the MTSLLNR model, combining local linear and truncated spline estimators.
• A modified weighted least square (WLS) method using two-stage estimation is employed.
• Optimal knots and bandwidth are selected using the generalized cross-validation (GCV) method.

Main Results:

• A simulation study with varying sample sizes and time points was conducted.
• The MTSLLNR model was applied to real-world poverty gap index data.
• Both simulation and real data analyses indicated the model's consistency and effectiveness.

Conclusions:

• The MTSLLNR model effectively models longitudinal data with diverse predictor variable patterns.
• The method shows good performance and consistency, validated by simulations and a real data application.
• The MTSLLNR model offers a robust approach for complex longitudinal data analysis.