A proportional hazards model for arbitrarily censored and truncated data

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Turnbull's nonparametric estimation method requires modification for both truncation and censoring.
• Existing methods struggle with simultaneously handling censored and truncated data.
• Accurate statistical modeling is crucial for understanding disease progression and induction times.

Purpose of the Study:

• To present a corrected and extended Turnbull's method for nonparametric estimation.
• To develop a proportional hazards model for arbitrarily censored and truncated data.
• To enable partial testing for zero regression coefficients using likelihood ratio or Wald tests.

Main Methods:

• Correction and extension of Turnbull's method for handling censored and truncated data.
• Development of a proportional hazards regression model.
• Application of likelihood ratio tests and Wald tests for coefficient testing.

Main Results:

• A robust method for nonparametric estimation with both censoring and truncation is presented.
• The proportional hazards model effectively handles complex survival data.
• The methodology allows for statistically significant testing of regression coefficients.

Conclusions:

• The corrected Turnbull's method and extended regression analysis provide a powerful tool for survival data.
• This approach is valuable for analyzing time-to-event data in medical research, such as AIDS induction and diabetic nephropathy.
• The developed statistical framework enhances the analysis of arbitrarily censored and truncated data.