NEWS.md
In survstan: Fitting Survival Regression Models via 'Stan'

survstan 0.0.1

Implementation of the Yang and Prentice (YP) model for dealing with survival data with possibly crossing survival curves.
Implementation of survfit.survstan method for all implemented models.
Improvement in the output of the anova.survstan function: now the chosen regression model and baseline distribution are displayed in the ANOVA table.
Addition of dist argument in the model fitting functions (aftreg(), ahreg(), phreg(), poreg() and ypreg()) as an alternative way to specify the baseline distribution (for compability with the survival::survreg() function).
Addition of init argument in the model fitting functions (aftreg(), ahreg(), phreg(), poreg() and ypreg()) with default value equals to 0 to fix the initial starting values when calling the rstan::optimizing() function.
Addition of extractAIC.survstan method required in the stats::step() function for carrying out model selection via stepwise procedures.
Addition of logLik.survstan method to extract the log-likelihood value of a fitted model.
AIC.survstan method now returns a data.frame with the aic values along with the corresponding number of model parameters when more than one fit is passed to the method.
Addition of gastric and ipass datasets.

The Weibull distribution is now the default baseline distribution (replacing the exponential distribution in previous versions).
Implementation of Birnbaum-Saunders (fatigue) distribution.
Implementation of the function rank_models to rank survstan models according to their AICs.

Inclusion of emmeans method for survstan models.
Implementation of gamma and Rayleigh distributions.
Implementation of se method for computation of standard errors.
Correction of covariance matrix for positive parameters.
Application of delta method to avoid negative lower bounds in confidence intervals for positive parameters.
Bug correction in the implementation of survival functions.

update summary output.
tidy.survstan method was reimplemented to follow broom stantards, and survstan::tidy() no long masks broom::tidy().
cross_time() function now allows for parallel computing.
reparametrization of AH model (see details in https://fndemarqui.github.io/survstan/).
implementation of extended hazard (EH) family of survival regression models.

All implemented models now can accommodate an offset variable in the linear predictors.
Implementation of Gompertz and generalized gamma (original Stacy's parametrization and alternative Prentice's parametrization) distributions.

survfit.survstan method now returns a data.frame with the observed survival times and their corresponding estimated baseline survivals when newdata = NULL.