Right-Censored Time Series Modeling by Modified Semi-Parametric A-Spline Estimator

Area of Science:

• Statistics
• Time Series Analysis
• Semiparametric Regression

Background:

• Semiparametric regression models time series data with censored observations.
• Traditional methods like Kaplan-Meier can distort data structure, especially with heavy censoring.
• Accurate estimation of model components is challenging with censored data.

Purpose of the Study:

• To develop a modified semiparametric estimator using adaptive splines (A-splines) for right-censored time series data.
• To overcome data irregularity and information loss issues inherent in traditional synthetic data approaches.
• To evaluate the performance of the proposed A-spline estimator against a benchmark B-spline estimator.

Main Methods:

• Utilized adaptive spline (A-spline) fitting for semiparametric regression.
• Introduced a modified semiparametric estimator to handle right-censored observations.
• Employed a B-spline estimator as a benchmark for comparison.
• Conducted Monte Carlo simulations and analyzed a real-world dataset.

Main Results:

• The proposed A-spline estimator effectively handles data irregularity caused by censoring.
• The method demonstrates reduced information loss compared to traditional synthetic data approaches.
• Performance evaluation showed the A-spline estimator's viability against the B-spline benchmark.

Conclusions:

• The modified A-spline approach offers a robust solution for semiparametric regression with right-censored time series data.
• This method provides a more accurate and less distorting alternative to existing techniques.
• The findings support the practical application of the A-spline estimator in analyzing censored time series.