Semiparametric inference for the scale-mixture of normal partial linear regression model with censored data

Area of Science:

• Statistics
• Econometrics
• Data Science

Background:

• Classical linear regression struggles with non-normal errors and nonlinearity.
• Censored data analysis requires robust statistical models.
• Existing methods are sensitive to outliers and heavy tails.

Purpose of the Study:

• To propose a robust partial linear regression model for censored data.
• To address departures from normality and nonlinearity simultaneously.
• To accommodate heavy tails and outliers in regression analysis.

Main Methods:

• Utilizing scale-mixture of normal (SMN) distributions for error modeling.
• Implementing B-spline approximation for flexibility.
• Developing an EM-type algorithm for maximum likelihood (ML) parameter estimation.

Main Results:

• The proposed SMN-based partial linear regression model demonstrates flexibility.
• Simulation studies confirm the model's robustness with heavy-tailed data.
• The methodology effectively handles censored data with non-ideal error distributions.

Conclusions:

• The SMN-based partial linear regression offers a powerful alternative for censored data.
• The model provides improved parameter estimation in the presence of outliers and non-normality.
• The approach is validated through simulations and real-world data analysis.