v1.5.1

Supplemented example code to assist with quick start to the package

Full Changelog: v1.5.0...v1.5.1

v1.5.0

Major changes (since 1.4.0)

• changed Cox model convergence checking to use eps parameter and deprecated the gtol parameter. This will not induce breaking changes but may slightly change numerical output. gtol was a convergence check using the gradient vector, which would often result in very slow- or non- convergence in large datasets. eps now follows the standard in the survival::coxph function in R, which checks tolerance by relative changes in the log partial likelihood (i.e. convergence is declared when abs(1-(LPL(i)-LPL(i-1)) / LPL(i-1)) < eps.

Major changes (since pre-release)

• Robust, jacknife, and bootstrap variance implemented
• Interface with optimizer for parametric survival regression models
• Distributions added for parametric survival regression models (generalized gamma)
• Predict method for parametric survival regression
• baseline hazard estimation
• Changed signatures for non-parametric methods to allow @formula

Minor changes (since pre-release)

• Documentation additions

Bug fixes (since pre-release)

• corrected p-value calculation from Cox model
• Aalen-Johansen bug with weights fixed
• coeftable bug with one predictor

v0.13.2

First pre-release. Most of package core functionality is in place and API is semi-stable

Features:

• Cox model
• Aalen-Johansen
• Kaplan-Meier
• Late entry
• Right censoring
• use of @formula interface from StatsAPI

Still in the works:

• Robust variance estimator
• Interface with Optimizer if default Newton-Raphson approach does not work