A FORTRAN program for multivariate survival analysis on the personal computer

Area of Science:

• Biostatistics
• Survival Analysis
• Computational Statistics

Background:

• Multivariate survival analysis is crucial for understanding complex failure processes.
• Competing risks scenarios, where multiple events can occur, require specialized analytical approaches.
• Existing models like Cox's proportional hazards model may have limitations in certain applications.

Purpose of the Study:

• To present a FORTRAN program for multivariate survival regression in competing risks.
• To model failure rates using log-linear functions of explanatory variables.
• To provide an efficient alternative to Cox's proportional hazards model.

Main Methods:

• Development of a FORTRAN program for competing risks survival analysis.
• Modeling the relevant failure rate as a log-linear function of explanatory variables.
• Utilizing the Newton-Raphson method for maximum likelihood estimation of coefficients.
• Incorporating time-dependent variables and time itself for flexible distribution parameterization.

Main Results:

• The program facilitates multivariate survival or life table regression analysis.
• It models failure rates as log-linear functions, accommodating various explanatory variables.
• The approach allows for parameterizing time-to-failure distributions in Gompertz-like or Weibull-like ways.
• Maximum likelihood estimates are efficiently obtained via the Newton-Raphson method.

Conclusions:

• The presented FORTRAN program offers a robust tool for competing risks survival analysis.
• The log-linear modeling approach provides flexibility and efficiency compared to traditional methods.
• The program is suitable for personal computers and handles large datasets acceptably.